UNIVERSITY OF CALICUT

SCHOOL OF DISTANCE EDUCATION

M.Com
( I SEMESTER)
(2019 Admn Onwards)

CORE COURSE : MCM1C03

QUANTITATIVE TECHNIQUES
FOR BUSINESS DECISIONS

190603
QUANTITATIVE TECHNIQUES FOR
BUSINESS DECISIONS.

STUDY MATERIAL FIRST SEMESTER

CORE COURSE : MCM1C03

M.Com
(2019 Admn Onwards)

UNIVERSITY OF CALICUT
SCHOOL OF DISTANCE EDUCATION
Calicut University- PO, Malappuram,

Kerala, India - 673 635

QUANTITATIVE TECHNIQUES FOR BUSINESS DECISIONS
M.COM FIRST SEMESTER
CONTENTS

Chapter No. Description Page No.

1 Introduction to Quantitative Techniques 1-7

2 Correlation Analysis 8 - 23

3 Regression Analysis 24 - 37

4 Probability Distributions 38 - 40

5 Binomial Distribution 41 - 45

6 Poisson Distribution 46 - 49

7 Normal Distribution 50 - 56

8 Exponential Distribution 57

9 Uniform Distribution 58

10 Statistical Inferences 59 - 76

11 Chi-Square Test 77 - 87

12 Analysis of Variance 88 – 99

13 Non-parametric Tests 100 – 114

14 Sample Size Determination 115 – 117

15 Statistical Estimation 118 – 120

16 Softwares for Quantitative Methods 121- 128

 1

CHAPTER 1

INTRODUCTION TO QUANTITATIVE TECHNIQUES

Quantitative technique is a very powerful tool with the help of which, the
business organizations can augment their production, maximize profits, minimize costs,
and production methods can be oriented for the accomplishment of certain pre –
determined objectives. Quantitative techniques are used to solve many of the problems that
arise in a business or industrial area. A large number of business problems, in the
relatively recent past, have been given a quantitative representation with considerable
degree of success. All this has attracted the business executives, public administrators alike
towards the study of these techniques more and more in the present times.
Managerial activities have become complex and it is necessary to make right
decisions to avoid heavy losses. Whether it is a manufacturing unit, or a service
organization, the resources have to be utilized to its maximum in an efficient manner. The
future is clouded with uncertainty and fast changing, and decision- making – a crucial
activity – cannot be made on a trial-and-error basis or by using a thumb rule approach. In
such situations, there is a greater need for applying scientific methods to decision-making to
increase the probability of coming up with good decisions. Quantitative Technique is a
scientific approach to managerial decision-making. The successful use of Quantitative
Technique for management would help the organization in solving complex problems on time,
with greater accuracy and in the most economical way.
Definitions
Since Quantitative technique is a practical methodological technique, there is
no precise definition for the term. Quantitative techniques are defined as “those
statistical techniques which lead to numerical analysis of variables, affecting a
decision situation, and evaluation of alternative strategies to attain objectives of
organizations.”
Quantitative techniques involves “ transformation of a qualitative
description of a decision situation, into quantitative format, identifying of
variables, setting out alternative solutions and supplementing decision making, by
replacing judgment and intuition.”
Quantitative techniques may be described as those techniques “which
provide decision maker, with a systematic and powerful tool of analysis, based on
quantitative and numeric data relating to alternative option.”
Thus quantitative techniques are a set of techniques involving numerical
formulation of a decision situation and analysis of variables, so as to arrive at
alternative solutions, leading to optimal decision.
Classification of quantitative techniques
Quantitative techniques are a set of methods used to quantitatively formulate,
analyze, integrate and decide problems or issues. They are broadly classified into three –
mathematical techniques, statistical techniques and programming techniques.
 2

Mathematical techniques
They are quantitative techniques in which numerical data are used along with the
principles of mathematics such as integration, calculus etc. They include permutations,
combinations, set theory, matrix analysis, differentials integration etc.
Permutations and combinations
Permutation is mathematical device of finding possible number of arrangements
or groups which can be made of a certain number of items from a set of observations. They
are groupings considering order of arrangements.
Combinations are number of selections or subsets which can be made of a certain
number of items from a set of observations, without considering order. Both combinations
and permutations help in ascertaining total number of possible cases.
Set theory
It is a modern mathematical device which solves the various types of critical
problems on the basis of sets and their operations like Union, intersection etc.
Matrix Algebra
Matrix is an orderly arrangement of certain given numbers or symbols in rows and
columns. Matrix analysis is thus a mathematical device of finding out the results of
different types of algebraic operations on the basis of relevant matrices. This is useful to
find values of unknown numbers connected with a number of simultaneous equations.
Differentials
Differential is a mathematical process of finding out changes in the dependent
variable with reference to a small change in the independent variable. It involves
differential coefficients of dependent variables with or without variables.
Integration
It is a technique just reversing the process of differentiation. It involves the
formula f(x) dx where f(x) is the function to be integrated
Statistical techniques
They are techniques which are used in conducting statistical inquiry concerning a
certain phenomenon. They include all the statistical methods beginning from the collection
of data till interpretation of those collected data. Important statistical techniques include
collection of data, classification and tabulation, measures of central tendency, measures of
dispersion, skewness and kurtosis, correlation, regression, interpolation and extrapolation,
index numbers, time series analysis, statistical quality control, ratio analysis , probability
theory, sampling technique, variance analysis, theory of attributes etc.
Programming techniques
These techniques focus on model building, and are widely applied by decision
makers relating to business operations. In programming, problem is formulated in
numerical form, and a suitable model is fitted to the problem and finally a solution is
derived. Prominent programming techniques include linear programming, queuing theory,
inventory theory, theory of games, decision theory, network programming, simulation,
replacement non linear programming, dynamic programming integer programming etc.
 3

Functions of Quantitative Techniques:

The following are the important functions of quantitative techniques:

1. To facilitate the decision-making process
2. To provide tools for scientific research
3. To help in choosing an optimal strategy
4. To enable in proper deployment of resources
5. To help in minimizing costs
6. To help in minimizing the total processing time required for performing a set of jobs

Quantitative and qualitative approaches

Decision making is the process of selecting optimal alternative from among
several alternatives, subject to states of nature. While analyzing a situation for such
a selection, two approaches can be adopted – quantitative approach and qualitative
approach
Quantitative approach
This approach involves generation and analysis of data in numerical form.
Data obtained as per quantitative approach can be subjected to rigorous quantitative
analysis in a formal fashion. This will reveal almost all inherent characteristics of
the variable under study.
Quantitative approach may further be subdivided into inferential, experimental and
simulation approaches. The purpose of inferential approach is to form a data base to infer
characteristics or relationships of variables. Required data would be usually obtained
through field survey.
Experimental approach is characterized by much greater control over the study
environment, and in this case variables are manipulated to observe their effect on other
variables.
Simulation approach involves the construction of an artificial environment or
model within which relevant information and data can be generated. This permits an
observation of dynamic behavior of the system or sub system under modeled conditions.
The term simulation, in the context of business, means building of a model, that represents
the structure of a dynamic process or operation.
Qualitative approach
Qualitative approach is concerned with subjective assessment of attitudes, opinions
and behavior. Decision making in such situations is a function of decision maker’s insight
and impressions. Such an approach generates results either in non-quantitative form or in a
form which cannot be subjected to rigorous quantitative analysis. For example, opinion that
a person may be good or bad
Basically, the techniques of focus group interviews, projective techniques and depth
interviews use qualitative approach for decision making.
 4

Generally there are four non quantitative techniques of decision making

Intuition – decision making on intuition is characterized by inner feelings of the
decision maker. It is purely subjective
Facts –It follows the rule that decision should be based on facts, and not on feelings.
Experiences – Experience is the most valuable asset, if used logically. Decisions
should be based on precedence.
Opinion – in decision making, expert opinions can be relied on. In fact, this is
widely used by all levels of managers.
However, even qualitative approach may be transformed into quantitative form, in
practical studies. This is achieved through measurement and scaling. Measurement is
assigning numbers or values to concepts or phenomena. Scaling refers to placing a
concept or characteristic on the appropriate position of a measured scale. For example,
Marital status of a person may be : (single)1 , (married)2, (divorced)3 (widowed)4. Here
qualitative or non quantitative data is logically converted into quantitative data.
Significance of quantitative decisions
Quantitative Techniques have proved useful in tackling managerial decision
problems relating to business and industrial operations. Quantitative decisions are
considered significant on the following grounds.
Simplifies decision making
Quantitative techniques simplify the decision making process. Decision theory
enables a manager to select the best course of action. Decision tree technique refines
executive judgment in systematic analysis of the problem, these techniques permit
scientific decision making under conditions of risk and uncertainty. Decision problems such
as manpower planning ,demand forecasting, selection suppliers, production capacities, and
capital requirements planning can be more effective tackled using quantitative techniques.
Scientific analysis
It provides a basis for precise analysis of the cause and effect relationship. They
make it possible to measure the risks inherent, in business by providing an analytical and
objective approach. These techniques reduce the need for intuition and objective approach.
In this way quantitative techniques enable managers to use logical thinking in the analysis
of organizational problems,
Allocation of resources
They are very helpful in the optimum deployment of resources. For example,
Programme Evaluation and Review Techniques enable a manager to determine the earliest
and the latest times for each of the events and activities involved in a project. The
probability of completing the project by a specified date can be determined. Timely
completion of the project helps to avoid time and cost overruns. Similarly, linear
programming technique is very useful in optimal allocation of scarce resources,
production scheduling and in deciding optimal assignments.
Profit maximization
Quantitative techniques are invaluable in assessing the relative profitability of
alternative choices and identifying the most profitable course of action. What should be the
relative mix of different products, which site to choose for location out of alternative sites,
 5

which arrangement of orders in terms of time and quantity, will give maximum profits.
Such question can be answered with the help of quantitative techniques.
Cost minimization
Quantitative techniques are helpful in tackling cost minimization problems. For
example waiting line theory enables a manager to minimize waiting and servicing costs.
Their techniques help business managers in taking a correct decision through analysis of
feasibility of adding facilities.
Forecasting
Quantitative techniques are useful in demand forecasting. They provide a scientific
basis of coping with the uncertainties of future demand. Demand forecasts serve as the
basis for capacity planning. Quantitative technique enables a manager to adopt the
minimum risk plan.
Inventory control
Inventory planning techniques help in deciding when to buy and how much to buy.
It enables management to arrive at appropriate balance between the costs and benefits of
holding stocks. The integrated production models technique is very useful in minimizing
costs of inventory, production and workforce. Statistical quality controls help us to
determine whether the production process is under control or not.
Applications of quantitative techniques in business operations
Quantitative techniques are widely applied for solving decision problems of routine
operations of business organizations. It is especially useful for business managers,
economist, statisticians, administrators, technicians and others in the field of business,
agriculture, industry services and defense. It has specific applications in the following
functional areas of business organizations.
Planning
In planning, quantitative techniques are applied to determine size and location of plant,
product development, factory construction, installation of equipment and machineries etc.
Purchasing
Quantitative techniques are applied in make or buy decisions, vendor
development, vendor rating, purchasing at varying prices, standardization and variety
reduction, logistics management.
Manufacturing
Quantitative techniques address questions like product mix, production planning,
quality control, job sequencing, and optimum run sizes.
Marketing
Marketing problems like demand forecasting, pricing competitive strategies,
optimal media planning and sales management can be solved through application
appropriate quantitative techniques.
Human resource management
Quantitative techniques supports decision making relating to man power planning
with due consideration to age, skill, wastage and recruitment , recruitment on the basis of
proper aptitude, method study , work measurement, job evaluation, development of
incentive plans, wage structuring and negotiating wage and incentive plan with the union.
 6

Research and Development

Quantitative techniques are helpful in deciding research issues like market
research, market survey, product innovation, process innovations, plant relocation, merger
and acquisitions etc.
Classification of quantitative techniques
Quantitative techniques are a set of methods used to quantitatively formulate,
analyze, integrate and decide problems or issues. They are broadly classified into three –
mathematical techniques, statistical techniques and programming techniques.
Mathematical techniques
They are quantitative techniques in which numerical data are used along with the
principles of mathematics such as integration, calculus etc. They include permutations,
combinations, set theory, matrix analysis, differentials integration etc.
Permutations and combinations
Permutation is mathematical device of finding possible number of arrangements
or groups which can be made of a certain number of items from a set of observations. They
are groupings considering order of arrangements.
Combinations are number of selections or subsets which can be made of a certain
number of items from a set of observations, without considering order. Both combinations
and permutations help in ascertaining total number of possible cases.
Set theory
It is a modern mathematical device which solves the various types of critical
problems on the basis of sets and their operations like Union, intersection etc.
Matrix Algebra
Matrix is an orderly arrangement of certain given numbers or symbols in rows and
columns. Matrix analysis is thus a mathematical device of finding out the results of
different types of algebraic operations on the basis of relevant matrices. This is useful to
find values of unknown numbers connected with a number of simultaneous equations.
Differentials
Differential is a mathematical process of finding out changes in the dependent
variable with reference to a small change in the independent variable. It involves
differential coefficients of dependent variables with or without variables.
Integration
It is a technique just reversing the process of differentiation. It involves the
formula f(x) dx where f(x) is the function to be integrated
Statistical techniques
They are techniques which are used in conducting statistical inquiry concerning a
certain phenomenon. They include all the statistical methods beginning from the collection
of data till interpretation of those collected data. Important statistical techniques include
collection of data, classification and tabulation, measures of central tendency, measures of
dispersion, skewness and kurtosis, correlation, regression, interpolation and extrapolation,
index numbers, time series analysis, statistical quality control, ratio analysis , probability
theory, sampling technique, variance analysis, theory of attributes etc.
Programming techniques
These techniques focus on model building, and are widely applied by decision
 7

makers relating to business operations. In programming, problem is formulated in

numerical form, and a suitable model is fitted to the problem and finally a solution is
derived. Prominent programming techniques include linear programming, queuing theory,
inventory theory, theory of games, decision theory, network programming, simulation,
replacement non linear programming, dynamic programming integer programming etc.
Quantification of qualitative data
In most cases, information is born in the form of qualitative description of
situations. This may be quantified. Such quantification leads to following favorable out
comes
1. It attracts readers’ attention to patterns in the information
2. It helps to memorize and stacking of information
3. It assists in timely retrieval of data.
4. It supports efficient decision making.

Limitations of Quantitative Techniques:

Even though the quantitative techniques are inevitable in decision-making process, they are not
free from short comings. The following are the important limitations of quantitative techniques:
1. Quantitative techniques involves mathematical models, equations and other mathematical
expressions
2. Quantitative techniques are based on number of assumptions. Therefore, due care must be
ensured while using quantitative techniques, otherwise it will lead to wrong conclusions.
3. Quantitative techniques are very expensive.
4. Quantitative techniques do not take into consideration intangible facts like skill, attitude
etc.
5. Quantitative techniques are only tools for analysis and decision-making. They are not
decisions itself.

REVIEW QUESTIONS:
1. Define Quantitative Techniques.
2. Explain the classification of quantitative techniques.
3. Explain the significance of quantitative decisions.
4. What are the uses of quantitative techniques in Business?
5. Explain the qualitative approach in decision making.
6. What are the important limitations of quantitative techniques?
 8

Chapter 2

CORRELATION ANALYSIS
Meaning and Definition of Correlation

Correlation analysis is an attempt to examine the relationship between two variables. It

analyses the association between two or more variables. It is a bi-variate analysis.

According to Croxton and Cowden, “when the relationship is of quantitative nature, the
appropriate statistical tool for discovering and measuring the relationship and expressing it in a brief
formula is known as correlation”

According to A.M.Tuttle, “Correlation is an analysis of the co-variation between two or more

variables.”

According to Ya-Lun-Chou, “Correlation analysis is an attempt to determine the degree of

relationship between variables.”

Correlation analysis helps to know the direction of relationship as well as the degree of
relationship exists between two or more variables.

Types of Correlation

I. Positive and Negative Correlation

(a) Positive Correlation: If two variables move in the same direction, then the correlation is
called positive. For example, price and supply are positively correlated. When price goes up,
supply goes up and vice versa.
(b) Negative Correlation: If two variables move in the opposite direction, then the correlation is
called negative. For example, price and demand are negatively correlated. When price goes up,
demand falls down and vice versa.

II. Simple, Partial and Multiple Correlation

(a) Simple Correlation: In a correlation analysis, if there are only two variables, then the
correlation analysis is called simple correlation. For example, the relationship between weight
and height, price and demand, price and supply, etc.
(b) Partial Correlation: When there are more than two variables and we study the relationship
between any two variables only, assuming other variables as constant, it is called partial
correlation. For example, the study of the relationship between rainfall and agricultural
produce, without taking into consideration the effects of other factors such as quality of seeds,
quality of soil, use of fertilizer, etc.
(c) Multiple Correlation: When there are more than two variables and we study the relationship
between one variable and all the other variables taken together, then it is the case of multiple
 9

correlation. Suppose there are three variables, namely x, y and z. The correlation between x
and (y & z) taken together is multiple correlation. Similarly, the relation between y and (x & z)
taken together is multiple correlation. Again, the relation between z and (x & y) taken together
is multiple correlation.

III. Linear and Non-linear Correlation

(a) Linear Correlation: When the amount of change in one variable leads to a constant
ratio of change in the other variable, the relationship is called linear correlation. For
example, if price falls down by 10%, it leads to a fall in supply by 12% each time, it is
linear correlation. When we plot the data on graph, we will get a straight line. Here, the
relationship between the variables may be expressed in the form of y = ax + b.
(b) Non-linear Correlation: When the amount of change in one variable does not lead to
a constant ratio of change in the other variable, the relationship is called non-linear
correlation. When we plot the data on graph, we never get a straight line. Therefore,
non-linear correlation is also called curvi-linear correlation.
IV. Logical and Illogical Correlation
(a) Logical Correlation: When the correlation between two variables is not only
mathematically defined but also logically sound, it is called logical correlation. For
example, correlation between price and demand.
(b) Illogical Correlation: When the correlation between two variables is mathematically
defined but not logically sound, it is called illogical correlation. For example,
correlation between availability of rainfall and height of people. This type of correlation
is also known as Spurious correlation or Non-sense correlation.

Methods of Studying Correlation

The various methods for studying correlation can be classified into two categories. They are:

I. Graphic Methods:
(1) Scatter diagram method
(2) Correlation Graph method
II. Mathematical Methods:
(1) Karl Pearson’s Product Moment Method
(2) Spearman’ Rank Correlation Method
(3) Concurrent Deviation Method

Graphic Methods:

Scatter Diagram Method

This is a simple method for analysing correlation between two variables. One variable is shown
on the X- axis and the other on the Y-axis. Each pair of values is shown on the graph paper
using dots. When all the pairs of observations are plotted as dots, the relationship exists
 10

between the variables is analysed by observing how the dots are scattered. If the dots show an
upward or downward trend, then the variables are correlated. We may interpret the scatter
diagram as follows:

(a) If all the dots are lying on a straight line from left bottom corner to the right upper corner,
there is perfect positive correlation between variables.
(b) If all the dots are lying on a straight line from left upper corner to the right bottom corner,
there is perfect negative correlation between variables.
(c) If all the dots are plotted on a narrow band from left bottom corner to the right upper
corner, there is high degree of positive correlation between variables.
(d) If all the dots are plotted on a narrow band from left upper corner to the right bottom
corner, there is high degree of negative correlation between variables.
(e) If all the dots are plotted on a wide band from left bottom corner to the right upper corner,
there is low degree of positive correlation between variables.
(f) If all the dots are plotted on a wide band from left upper part to the right bottom part, there
is low degree of negative correlation between variables.
(g) If the plotted dots do not show any trend, the variables are not correlated.

Correlation Graph Method

In correlation graph method, separate curves are drawn for each variable on the same
graph. The relationship between the variables is interpreted on the basis of the direction and
closeness of the curves. If both the curves move in the same direction, there is positive
correlation and if they are moving in opposite directions, there is negative correlation between
the variables.

Mathematical Methods:

Under mathematical methods, the correlation between variables is studied with the help of a
numerical value obtained using an appropriate formula. This numerical value is called
coefficient of correlation. Coefficient of correlation explains both the direction as well as
degree of relationship exists between the variables.

Degree of Correlation:

The degree of correlation can be classified as follows:

(a) Perfect Positive Correlation: When coefficient of correlation is +1

(b) Perfect Negative Correlation: When coefficient of correlation is –1
(c) High Degree of Positive Correlation: When coefficient of correlation lies between + 0.75
and + 1
(d) High Degree of Negative Correlation: When coefficient of correlation lies between –
0.75 and –1
(e) Moderate Degree of Positive Correlation: When coefficient of correlation lies between +
0.5 and + 0.75
 11

(f) Moderate Degree of Positive Correlation: When coefficient of correlation lies between –
0.5 and – 0.75
(g) Low Degree of Positive Correlation: When coefficient of correlation lies between 0 and +
0.33
(h) Low Degree of Negative Correlation: When coefficient of correlation lies between 0 and
– 0.33
(i) No Correlation: When coefficient of correlation is zero.

Karl Pearson’s Product Moment Method

This is the popularly used method for analysing correlation. This method is designed by a
reputed Statistician, Prof. Karl Pearson and therefore, it is generally known as Pearsonian
Coeffiecient of Correlation. Karl Pearson’s coefficient of correlation is denoted by ‘r’ Under
this method, coefficient of correlation is computed by usin any one of the following formulae:

εxy
r =
√εx2 εy2
where x = deviations of X values from its actual mean
y = deviations of X values from its actual mean

OR
nεdxdy – (εdx . εdy)
r =
√nεdx2 – (εdx)2 nεdy2 – (εdy)2
where n = number of pairs of observations
dx = deviations of observations (variable x) from assumed mean
dy = deviations of observations (variable y) from assumed mean

OR
nεXY – (εX . εY)
r =
√nεX2 – (εX)2 nεY2 – (εY)2
where n = number of pairs of observations
X = Given values of variable X
Y = Given values of variable Y

Qn: From the following data compute product moment correlation coefficient and interpret it:

X 57 42 40 38 42 45 42 44 40 46 44 43
Y 10 26 30 41 29 27 27 19 18 19 31 29

Sol:
 12

Computation of Product Moment Correlation Coefficient

X Y dx = (x --45) dy=(y—25) Dxdy dx2 dy2
57 10 12 -15 -180 144 225
42 26 -3 1 -3 9 1
40 30 -5 5 -25 25 25
38 41 -7 16 -112 49 256
42 29 -3 4 -12 9 16
45 27 0 2 0 0 4
42 27 -3 2 -6 9 4
44 19 -1 -6 6 1 36
40 18 -5 -7 35 25 49
46 19 1 -6 -6 1 36
44 31 -1 6 -6 1 36
43 29 -2 4 -8 4 16
εdx= -17 εdy= 6 εdxdy= -317 εdx2=277 εdy2 = 704

nεdxdy – (εdx . εdy)

r =
√nεdx2 – (εdx)2 nεdy2 – (εdy)2

r = (12 x -317) – (-17 x 6) / √[(12 x 277) – (-17)2] [(12 x 704) – (6)2]

= –3804 – –102 / [√3324 – 289] [8448 – 36]

= – 3702 / √3035 x 8412 = – 14208 / √25530420

= – 3702 / 5052.76 = – 0.733

There is a high degree of negative correlation between x and y.

Properties of Coefficient of Correlation

1. Coefficient of correlation lies between – 1 and + 1

2. It is a pure numerical value independent of the units of measurement.
3. It does not change with reference to the change of origin or change of scale.
4. It is the geometric mean of the two regression coefficients
5. It is computed by using a well defined formula.
6. Coefficient of correlation between x and y and that between y and x are same.

Probable Error
Probable error is a statistical device used to measure the reliability and dependability of
the value of correlation coefficient. When the numerical value of probable error is added to and
 13

subtracted from the value of correlation coefficient, we get two limits within which the
population parameter is expected to lie.

Probable Error (P.E) = 0.6745 x Standard Error

Probable Error (P.E) = 0.6745 x [(1 – r2)/ √n]
where n= number of pairs of observations; r = correlation coefficient.

Uses of Probable Error

1. Probable error can be used to measure the reliability and dependability of coefficient of
correlation
2. It helps to determine the limits within which population parameter is expected to lie.
3. With the help of P.E, coefficient of correlation can be interpreted more accurately:
(a) If ‘r’ is less than P.E, there is no evidence of correlation.
(b) If ‘r’ is more than 6 times of P.E, correlation is significant
(c) If ‘r’ is 0.5 or more and the P.E is not much, the correlation is considered to be
significant.

Qn: Find Standard Error (S.E) and Probable Error (P.E), if r = 0.8 and number of pairs of observations
= 64. Also interpret the value of ‘r’.

Sol:

Standard Error (S.E) = 1– r2/ √n

= 1– 0.82/ √64 = 1– 0.64/8

= 0.36/8 = 0.045

Probable Error (P.E) = 0.6745 x Standard Error

= 0.6745 x 0.045 = 0.0304

r/P.E = 0.08/0.0304 = 26.32.

Since ‘r’ is more than 26.32 times of P.E, the value of ‘r’ is highly significant.

Qn: Following table shows the marks obtained by students in two courses:

Course I 45 70 65 30 90 40 50 75 85 60
Course II 35 90 70 40 95 40 60 80 80 50
Find coefficient of correlation and P.E. Is ‘r’ significant?

Sol:

Computation of Product Moment Correlation Coefficient

Course I Course dx = (x --60) dy=(y—60) Dxdy dx2 dy2
(x) II (y)
 14

45 35 -15 -25 375 225 625

70 90 10 30 300 100 900
65 70 5 10 50 25 100
30 40 -30 -20 600 900 400
90 95 30 35 1050 900 1225
40 40 -20 -20 400 400 400
50 60 -10 0 0 100 0
75 80 15 20 300 225 400
85 80 25 20 500 625 400
60 50 0 -10 0 0 100
εdx= 10 εdy= 40 εdxdy= εdx2= εdy2 =
3575 3500 4550

nεdxdy – (εdx . εdy)

r =
√nεdx2 – (εdx)2 nεdy2 – (εdy)2

r = (10 x 3575) – (10 x 40) / √[(10 x 3500) – (10)2] [(10 x 4550) – (40)2]

= 35750 – 400 / √[35000 – 100] [45500 – 1600]

= 35350 / √34900 x 43900 = 35350 / 39142.177

= + 0.903

Probable Error (P.E) = 0.6745 x [(1 – r2)/ √n]

= 0.6745 x [1 – 0.9032)/ √10]

= 0.6745 x [1 – 0.815)/ √10]

= 0.6745 x [0.185/ 3.162]

= 0.6745 x [0.0585] = 0.0395

r/P.E = 0.903/0.0395 = 22.86

Since coefficient of correlation is more than 22.86 of P.E, ‘r’ is very significant.

Coefficient of Determination

Coefficient of Determination is defined as the ratio of the explained variance to the

total variance. It denoted by r2 and is usually expressed as percentage. Coefficient of
Determination explains the percentage of the variation in the dependent variable that can be
explained in terms of the independent variable.
 15

Coefficient of Determination = r2

Coefficient of Determination = Explained Variance/Total Variance

Coefficient of non-determination = 1– r2.

Qn: If the coefficient of correlation between two variables is 0.85, what percentage of variation of
dependent variable is explained? Also find the coefficient of non-determination.

Sol:

Coefficient of Determination
(Percentage of explained Variance) = r2
= 0.85 x 0.85 = 0.7225 = 72.25%
Coefficient of non-determination = 1– r2 = 1 – 0.7225= 2775 = 27.75%

Rank Correlation Method

When the variables cannot be measured in quantitative terms, the coefficient of correlation can
be found out by using rank correlation method. Here ranks are to be assigned to the individual
observations. Ranks may be assigned in either ascending or descending order. This method was
designed by Charles Edward Spearman in 1904. He suggested two formulae for computing rank
correlation coefficient. Rank correlation coefficient if denoted by ‘R’
(1) When there is no equal rank:

6εD2
R =1 –
(n3 – n)

where D = Rank difference

n = Number of pairs of observations
(2) When there are equal ranks:

6{εD2 + [(m3-m)/12] + [(m3-m)/12] + ....}

R = 1 –
(n3 – n)

where D = Rank difference

n = Number of pairs of observations
m = Number of times a particular rank repeats
Qn: The ranks of 6 students in two courses are given below:
Rank for Course I 6 1 5 2 4 3
Ran for Course II 3 1 4 2 5 6
Compute Spearman’s Rank Correlation Coefficient.
Sol:
 16

Here there is no equal rank.

6εD2
R =1 –
(n3 – n)

Computation of Rank Correlation Coefficient

Rank for Course I Rank for Course I(R2) D ( R1 – R2) D2
(R1)
6 3 3 9
1 1 0 0
5 4 1 1
2 2 0 0
4 5 -1 1
3 6 -3 9
εD2 = 20
6εD2
R =1 –
(n3 – n)

= 1 – [(6 x 20)/(63–6)] = 1 – (120/210) = 1– 0.5714 = 0.4286

Qn: From the following data, compute Spearman’s Rank Correlation Coefficient:
x 330 332 328 331 327 325
y 415 434 420 430 424 428
Sol:
Here ranks are not given. So, at first, we have to assign ranks to each observation:

6εD2
R =1 –
(n3 – n)]

Computation of Rank Correlation Coefficient

X Y R1 R2 D (R1 – R2) D2
330 415 4 1 3 9
332 434 6 6 0 0
328 420 3 2 1 1
331 430 5 5 0 0
327 424 2 3 -1 1
325 428 1 4 -3 9
εD2 = 20
 17

6εD2
R =1 – = 1
(n3 – n)]
= 1 – [(6 x 20)/(63–6)] = 1 – (120/210) = 1– 0.5714 = 0.4286
Qn: From the following data, compute Spearman’s Rank Correlation Coefficient:

x 80 45 55 58 55 60 45 68 70 45 85
y 82 56 50 43 56 62 64 65 70 64 90
Sol:
This is the case of equal marks.

6{εD2 + [(m3-m)/12] + [(m3-m)/12] + ....}

R = 1 –
(n3 – n)

Computation of Rank Correlation Coefficient

X Y R1 R2 D (R1 – R2) D2
80 82 10 10 0 0
45 56 2 3.5 -1.5 2.25
55 50 4.5 2 2.5 6.25
58 43 6 1 5 25
55 56 4.5 3.5 1 1
60 62 7 5 2 4
45 64 2 6.5 -4.5 20.25
68 65 8 8 0 0
70 70 9 9 0 0
45 64 2 6.5 -4.5 20.25
85 90 11 11 0 0
εD2 = 79

6{79 + [(33-3)/12] + [(23-2)/12] + [(23-2)/12] + [(23-2)/12]}

R = 1 –
(113 – 11)

6 [79 + (2+0.5+0.5+0.5)]
=1 –
1320
= 1 – [6(79 + 3.5) ÷ 1320] = 1 – (495/1320) = 1– (0.375) = 0.625

Concurrent Deviation Method

This is a simple method for computing coefficient of correlation. Here, we consider only the
direction of change and not the magnitude of change. The coefficient of correlation is determined on
 18

the basis of number of concurrent deviations. That is why this method is named as such. The
coefficient of concurrent deviation is denoted by rc.
The formula for computing coefficient of concurrent deviation is:

rc = ± √± (2c – n) / n

where c = number of concurrent deviations

n = number of pairs of signs (not the pairs of observations)
Qn: Calculate coefficient of concurrent deviation from the following data:
x 180 182 186 191 183 185 189 196 193
y 246 240 230 217 233 227 215 195 200
Sol:

rc = ± √± (2c – n) / n
Computation of Coefficient of Concurrent Deviation
x y dx Dy Dxdy
180 246 ... ... ...
182 240 + - -
186 230 + - -
191 217 + - -
183 233 - + -
185 227 + - -
189 215 + - -
196 195 + - -
193 200 - + -
c = 0
Number of concurrent deviations = 0
rc = ± √± (2c – n) / n

rc = ± √± (2 x 0 – 8) / 8 = ±√±(0—8)/8 = –1
There is perfect negative correlation between x and y.

PARTIAL CORRELATION

When there are more than two variables and we study the relationship between any two
variables only, assuming other variables as constant, it is called partial correlation. For example, the
study of the relationship between rainfall and agricultural produce, without taking into consideration
the effects of other factors such as quality of seeds, quality of soil, use of fertilizer, etc.

Partial correlation coefficient measures the relationship between one variable and one of the
other variables assuming that the effect of the rest of the variables is eliminated.
 19

Suppose there are 3 variables namely x1, x2 and x3. Here, we can find three partial correlation
coefficients. They are:

(1) Partial Correlation coefficient between x1 and x2, keeping x3 as constant. This is denoted by
r12.3
(2) Partial Correlation coefficient between x1 and x3, keeping x2 as constant. This is denoted by
r13.2
(3) Partial Correlation coefficient between x2 and x3, keeping x1 as constant. This is denoted by
r23.1

The formulae for computing the above partial correlation coefficients are:

r12 – r13 r23

r12.3 =
√1–r132 √1–r232

r13 – r12 r23

r13.2 =
√1–r122 √1–r232

r23 – r12 r13

r23.1 =
√1–r122 √1–r132
Qn: If r12 = 0.98, r13 = 0.44 and r23 = 0.54, find (1) r12.3, (2) r13.2 and (3) r23.1

Sol:
(1)
r12 – r13 r23
r12.3 =
√1–r132 √1–r232

0.98 – (0.44 x 0.54)

r12.3 =
√1–0.442 √1–0.542

0.98 – 0.2376
=
√1–0.1936 √1–0.2916

= 0.7424/ (0.898 x 0.842) = 0.7424 / 0.7561 = 0.982

(2)
r13 – r12 r23
r13.2 =
 20

√1–r122 √1–r232

0.44 – (0.98 x 0.54)

r12.3 =
√1–0.982 √1–0.542

0.44 – 0.5292
=
√1–0.9604 √1–0.2916

= – 0.0892/ (0.199 x 0.842) = – 0.0892 / 0.1676 = – 0.5322

3)

r23 – r12 r13

r23.1 =
√1–r122 √1–r132

0.54 – (0.98 x 0.44)

r12.3 =
√1–0.982 √1–0.442

0.54 – 0.4312
=
√1–0.9604 √1–0.1936

= 0.1088/ (0.199 x 0.898) = 0.1088 / 0.1787 = 0.6088

MULTIPLE CORRELATION

When there are more than two variables and we study the relationship between one
variable and all the other variables taken together, then it is the case of multiple correlation.
Suppose there are three variables, namely x, y and z. The correlation between x and (y & z)
taken together is multiple correlation. Similarly, the relation between y and (x & z) taken
together is multiple correlation. Again, the relation between z and (x & y) taken together is
multiple correlation. In all these cases, the correlation coefficient obtained will be termed as
coefficient of multiple correlation.
Suppose there are 3 variables namely x1, x2 and x3. Here, we can find three multiple
correlation coefficients. They are:

1. Multiple Correlation Coefficient between x1 on one side and x2 and x3 together on the other
side. This is denoted by R1.23
2. Multiple Correlation Coefficient between x2 on one side and x1 and x3 together on the other
side. This is denoted by R2.13
3. Multiple Correlation Coefficient between x3 on one side and x1 and x2 together on the other
side. This is denoted by R3.12
 21

The formulae for computing the above multiple correlation coefficients are:

R1.23 = √ [r122 + r132 – 2 r12 r13 r23]÷ [1– r232]

R2.13 = √ [r122 + r232 – 2 r12 r13 r23]÷ [1– r132]

R3.12 = √ [r132 + r232 – 2 r12 r13 r23]÷ [1– r122]

Qn: If r12 = 0.6, r23 = r13 = 0.8, find R1.23, R2.13 and R3.12 .

Sol:

R1.23 = √ [r122 + r132 – 2 r12 r13 r23]÷ [1– r232]

= √[0.62 + 0.82 – 2 x 0.6 x 0.8 x 0.8]÷ [1– 0.82]

= √[0.36 + 0.64 – 0.768]÷ [1– 0.64]

= √0.232 /0.36 = √0.6444 = 0.8028

R2.13 = √ [r122 + r232 – 2 r12 r13 r23]÷ [1– r132]

= √[0.62 + 0.82 – 2 x 0.6 x 0.8 x 0.8]÷ [1– 0.82]

= √[0.36 + 0.64 – 0.768]÷ [1– 0.64]

= √0.232 /0.36 = √0.6444 = 0.8028

R3.12 = √ [r132 + r232 – 2 r12 r13 r23]÷ [1– r122]

= √[0.82 + 0.82 – 2 x 0.6 x 0.8 x 0.8]÷ [1– 0.62]

 22

= √[0.64 + 0.64 – 0.768]÷ [1– 0.36]

= √0.512 /0.64 = √0.8 = 0.8944

REVIEW QUESTIONS:

1. What do you mean by correlation analysis?

2. Define correlation.
3. What is scatter diagram? What are its advantages?
4. What is coefficient of correlation? What are its properties?
5. What are the different types of correlation?
6. What do you mean by degree of correlation?
7. What is meant by positive and negative correlation?
8. What are the merits of Karl Pearson’s Coefficient of Correlation?
9. What are the demerits of Karl Pearson’s Coefficient of Correlation?
10. What is rank correlation?
11. What are the merits of rank correlation?
12. What are the demerits of rank correlation?
13. What is meant by correlation graph method?
14. What is concurrent deviation method?
15. What is the main drawback of concurrent deviation method?
16. What is coefficient of determination?
17. What do you mean by coefficient of non-determination?
18. What do you mean by linear and non-linear correlation?
19. What is multiple correlations?
20. What is partial correlation?
21. What is probable error?
22. What are the uses of probable error?
23. From the following data, find coefficient of correlation and give interpretation:
X 200 270 400 310 340
Y 150 162 180 180 170
24. Find n, if P.E = 0.034 and r = 0.917
25. Find coefficient of correlation using Edward Spearman’s method:
Roll No. 1 2 3 4 5 6 7 8 9 10
 23

Marks I 45 56 39 54 45 40 56 60 30 35
Marks II 40 56 30 44 36 32 45 42 20 36

26. Compute coefficient of concurrent deviation:

X 100 110 110 120 122 125
Y 120 140 160 160 130 110
27. If r12 = 0.7, r13 = 0.61, r23 = 0.4, find r12.3, r13.2 and r 23.1
28. If r12 = 0.98, r13 = 0.44, r23 = 0.54, find R1.23, R2.13 and R 3.12
 24

Chapter 3

REGRESSION ANALYSIS
Meaning and Definition of Regression Analysis

Correlation analysis helps to know whether two variables are related or not. Once the
relationship between two variables is established, the same may be used for the purpose of predicting
the unknown value of one variable on the basis of the known value of the other. For this purpose we
have to examine the average functional relationship exists between the variables. This is known as
regression analysis.

Regression analysis may be defined as the process of ascertaining the average functional
relationship exists between variables so as to facilitate the mechanism of prediction or estimation or
forecasting. Regression analysis helps to predict the unknown values of a variable with the help of
known values of the other variable. The term regression was firstly used by Francis Galton.

Types of Regression

Regression may be classified as follows:

I. On the basis of number of variables:

(a) Simple Regression
(b) Multiple Regression
II. On the basis of Proportion of change in the variables:
(a) Liner Regression
(b) Non-liner Regression
1. Simple Regression

In a regression analysis, if there are only two variables, it is called simple regression analysis.

2. Multiple Regression

In a regression analysis, if there are more than two variables, it is called multiple regression analysis.

3. Linear Regression

In a regression analysis, if linear relation exists between variables, it is called linear regression
analysis. Under this, when we plot the data on a graph paper, we get a straight line. Here, the
relationship exists between variables can be expressed in the form of y = a + bx. In case of linear
regression, the change in dependent variable is proportionate to the changes in the independent
variable.

4. Non-linear Regression:
In case of non-linear regression, the relation between the variables cannot be expressed in the
form of y = a + bx. When the data are plotted on a graph, the dots will be concentrated, more
or less, around a curve. This is also called curvi-linear regression.
 25

Regression Line (Line of Best Fit)

Regression line is a graphical method to show the functional relationship between two
variables, namely dependent variable and independent variable. Since regression line helps to estimate
the unknown values of dependent variable, based on the known values of the independent variable, it
is also called estimating line (line of average).

According to Francis Galton, “The regression lines show the average relationship between two
variables.”

Regression Equations

Regression equations are algebraic expression of regression lines. As there are two regression
lines, there are two regression equations. They are:

(a) Regression Equation of X on Y : It shows the change in the value of variable X for a given
change in the value of variable Y.
(b) Regression Equation of Y on X : It shows the change in the value of variable Y for a given
change in the value of variable X.

Methods of drawing Regression Lines

There are two methods for drawing regression lines. They are:

(a) Free hand curve method

(b) The method of least squares.

Free hand Curve Method:

This is a simple method for constructing regression lines. Under this method, the values of paired
observations of the variable are plotted, by way of dots, on a graph paper. The X- axis represents the
independent variable and Y-axis represents the dependent variable. After observing how the dots are
scattered on the graph paper, we draw a straight line in such a way that the areas of the curve above
and below the line are approximately equal. The line so drawn clearly indicates the tendency of the
original data. Since there is subjectivity, this method is not commonly used in practice.

Qn: From the following data, draw a regression line of Y on X:

X 10 16 24 36 48
Y 20 12 32 40 55
Sol:
 26

Y
60

50

40

30

20

10

0
0 10 20 30 40 50 60

Method of Least Squares

Under method of least squares, the regression line should be drawn in such a way that the sum
of the squares of the deviations of the actual Y-values from the computed Y-values is the least. In
other words, 𝛆(y – yc)2 = minimum. The line so fitted is called line of best fit.

Methods of Calculating Regression Equations

The following are the two important methods for calculating regression equations:

1. Normal equation method

2. Regression coefficient method

Normal Equation Method

The General form of regression equation is:

X on Y : X = a + bY
Y on X : Y = a + bX

For finding out the constants ‘a’ and ‘b’, we have to develop and solve certain equations, called
normal equations. Therefore, this method is called normal equation method.

The normal equations computing ‘a’ and ‘b’ in respect of regression equation X on Y are:

εX = Na + bεY, and
εXY = aεY + bεY2
 27

The normal equations computing ‘a’ and ‘b’ in respect of regression equation Y on X are:

εY = Na + bεX, and
εXY = aεX + bεX2
After computing the values of the constants ‘a’ and ‘b’, substitute them to the respective
regression equations.

Qn: From the following data, fit the two regression equations:

x 4 5 8 2 1
y 5 6 7 3 2
Sol:

Regression Equation X on Y is:

X = a + bY

The normal equations to find the values of ‘a’ and ‘b’ are:

εX = Na + bεY, and

εXY = aεY + bεY2

Computation of Regression Equations

x y Xy x2 y2
4 5 20 16 25
5 6 30 25 36
8 7 56 64 49
2 3 6 4 9
1 2 2 1 4
εX=20 εY=23 εXY=114 εX2=110 εY2=123

20 = 5a + 23 b .............................. (1)

114 = 23a + 123 b ................................. (2)

(1) x 23 : 460 = 115 a + 529 b .......................... (1)
(2) X 5 : 570 = 115a + 615 b .......................... (2)
(3) -- (1) : 110 = 0 + 86 b

86 b = 110

b = 110/86 = 1.28

Substitute b= 1.2 in equation number (1)

20 = 5a + 23 x 1.28;
 28

20 = 5a + 29.44; 5a = 20 – 29.44; 5a = – 9.44

a = -9.44/5 = – 1.89

Substitute the values of ‘a’ and ‘b’ in regression equation X on Y:

X = – 1.89 + 1.28y

Regression Equation Y on X is:

Y = a + bX

The normal equations are:

εY = Na + bεX, and

εXY = aεX + bεX2

23 = 5a + 20b ............................. (1)

114 = 20a + 110b ............................. (2)

(1) x 4 : 92 = 20a + 80b ............................. (1)

114 = 20a + 110b ............................. (2)
(2) – (1): 22 = 0 + 30b
30b = 22
b = 22/30 = 0.73
Substitute b=0.73 in equation (1)

23 = 5a + 20 x 0.73; 23 = 5a + 14.6

5a = 23 – 14.6 = 8.4; a = 8.4/5 = 1.68

Substitute a = 1.68 and b = 0.73 in the general form of Y on X

Y = 1.68 + 0.73x

Regression Equation X on Y : X = – 1.89 + 1.28y

Regression Equation X on Y : Y = 1.68 + 0.73x

Regression coefficient method

Under regression coefficient method, regression equations are developed with the help of
regression coefficients. Since there are two regression equations, two regression coefficients are to be
computed.

The regression coefficient used to find the regression equation X on Y is “regression

Coefficient of X on Y”. It is denoted by bxy
 29

The regression coefficient used to find the regression equation Y on X is “regression

Coefficient of Y on X”. It is denoted by byx

The regression Equation X on Y is:

X – X̄ = bxy (Y – Ȳ )

Where X̄ = actual mean of X variable, Ȳ = actual mean of Y variable

bxy is computed by using any one of the following formula:

bxy = r . (σx/σy)

where bxy = Regression Coefficient of Regression equation X on Y

r = Coefficient of correlation
σX = Standard deviation of series X
σy = Standard deviation of series Y
OR

bxy = εxy/εy2

where bxy = Regression Coefficient of Regression equation X on Y

x = Deviation of X values from its actual mean
y = Deviation of X values from its actual mean
OR

nεdxdy – [(εdx) (εdy)]

bxy =
nεdy2 – (εdy)2

where bxy = Regression Coefficient of Regression equation X on Y

dx = Deviation X values from its assumed mean
dy = Deviation Y values from its assumed mean
n = Number of paired observations
OR

nεXY – [(εX) (εY)]

bxy =
nεY2 – (εY)2

where bxy = Regression Coefficient of Regression equation X on Y

X = Given values of X variable
Y = Given values of Y variable
 30

n = Number of paired observations

The regression Equation Y on X is:

Y – Ȳ = byx (X – X̄)

Where X̄ = Actual mean of X variable, Ȳ = Actual mean of Y variable

Byx is computed by using any one of the following formula:

byx = r . (σy/σx)

where byx = Regression Coefficient of Regression equation Y on X

r = Coefficient of correlation
σx = Standard deviation of series X
σy = Standard deviation of series Y
OR

byx = εxy/εx2

where byx = Regression Coefficient of Regression equation Y on X

x = Deviation of X values from its actual mean
y = Deviation of X values from its actual mean
OR

nεdxdy – [(εdx) (εdy)]

byx =
nεdx2 – (εdx)2

where byx = Regression Coefficient of Regression equation Y on X

dx = Deviation X values from its assumed mean
dy = Deviation Y values from its assumed mean
n = Number of paired observations
OR
 31

nεXY – [(εX) (εY)]

byx =
nεX2 – (εX)2

where byx = Regression Coefficient of Regression equation Y on X

X = Given values of X variable
Y = Given values of Y variable
n = Number of paired observations
Qn: You are given the following bivariate data:

X 7 2 1 1 2 3 2 6
Y 2 6 4 3 2 2 8 4
Using regression coefficients:

(a) Fit the regression equation of Y on X and predict Y if X = 5

(b) Fit the regression equation of X on Y and predict X if Y = 20

Sol.

The regression Equation Y on X is:

Y – Ȳ = byx (X – X̄)

Assumed mean method is used to find byx

nεdxdy – [(εdx) (εdy)]

byx =
nεdx2 – (εdx)2
Computation of Regression Equations
X Y dx (X–4) dy (Y–3) dxdy dx2 dy2
7 2 3 -1 -3 9 1
2 6 -2 3 -6 4 9
1 4 -3 1 -3 9 1
1 3 -3 0 0 9 0
2 2 -2 -1 2 4 1
3 2 -1 -1 1 1 1
2 8 -2 5 -10 4 25
6 4 2 1 2 4 1
𝛆X=24 𝛆Y=31 εdx = -8 εdy = 7 εdxdy= -17 εdx2 =44 εdy2= 39

8 x -17 – (-8 x 7)
byx =
8 x 44 – (-8)2
 32

= (-136 – -56) / (352 – 64) = -80/288 = – 0.278

X̄ = εX/n = 24/8 = 3
Ȳ = εY/n = 31/8 = 3.875

The regression Equation Y on X is:

Y – 3.875 = – 0.278 (X – 3)

Y = – 0.278 X + 0.834 + 3.875 = – 0.278 X + 4.709

Y = – 0.278 X + 4.709

If X=5, Y = (-0.278 x 5) + 4.709 = -1.39 + 4.709 = 3.319

The regression Equation X on Y is:

X – X̄ = bxy (Y – Ȳ )

nεdxdy – [(εdx) (εdy)]

bxy =
nεdy2 – (εdy)2
8 x -17 – (-8 x 7)
bxy =
8 x 39 – (7)2

= (-136 – -56) / (312 – 49) = -80/263 = – 0.3042

X̄ = εX/n = 24/8 = 3
Ȳ = εY/n = 31/8 = 3.875

The regression Equation X on Y is:

X– 3 = – 0.3042 (Y – 3.875)

X = – 0.3042 Y + 1.179 + 3 = – 0.3042Y + 4.179

X = – 0.3042Y + 4.179

If Y=20, X = (-0.3042 x 20) + 4.179 = -6.084 + 4.179 = –1.905

Properties of Regression Coefficients

1. In a bivariate data, there will be two regression coefficients. They are bxy and byx
2. bxy is the regression coefficient of regression equation X on Y
 33

3. byx is the regression coefficient of regression equation Y on X

4. Both the regression equations will have the same signs.
5. The sign of regression coefficients and correlation coefficient will be same.
6. The geometric mean of two regression coefficients is equal to coefficient of correlation.
√ bxy x byx = r
7. The product of two regression coefficients is equal to coefficient of determination.
bxy x byx = r2
8. When there is perfect correlation between X and Y, then bxy and x byx will be reciprocals of
each other.
9. When the standard deviations of both the variables are same, then the values of regression
coefficients and correlation coefficient will be same.
10. Both the regression coefficients will not be greater than 1. In other words, one of them can
be greater than 1; or both of them can be less than 1.

MULTIPLE REGRESSION

In multiple regression there are more than two variables. Here, we examine the effect of
two or more x3ndependent variables on one dependent variable. Suppose there are three
variables, namely, x1, x2 and x3. Here we may find three regression equations. They are:

1. Regression equation of x1 on x2 and x3

2. Regression equation of x2 on x1 and x3
3. Regression equation of x3 on x1 and x2

Equations of regression lines are generally termed as equations of planes of regression.

Following are the formulae for computing the above 3 regression plane equations:

1. Regression equation of x1 on x2 and x3:

(x1 – x̄1) = b12.3(x2 – x̄2) + b13.2(x3 – x̄3)

2. Regression equation of x2 on x1 and x3:

(x2 – x̄2) = b21.3(x1 – x̄1) + b23.1(x3 – x̄3)

3. Regression equation of x3 on x1 and x2:

(x3 – x̄3) = b31.2(x1 – x̄1) + b32.1(x2 – x̄2)

 34

where x̄1, x̄2 and x̄3 are actual means of x1, x2 and x3 respectively.

Yule’s Notation

Yule suggested that, the above equations may be simplified by taking (x3 – x̄3) = X1, (x3 – x̄3) = X2
and (x3 – x̄3) = X3. Then the equations of planes of regression are:

1. Regression equation of x1 on x2 and x3:

X1 = b12.3 X2+ b13.2 X3

2. Regression equation of x2 on x1 and x3:

X2 = b21.3 X1+ b23.1 X3

3. Regression equation of x3 on x1 and x2:

X3 = b31.2 X1+ b32.1 X2

In the above three equations, we used six regression coefficients. Following are the formulae
for computing regression coefficients:

b12.3 = (σ1/ σ2 ) [(r12 – r13r23)/(1– r232)]

b13.2 = (σ1/ σ3 ) [(r13 – r12r23)/(1– r232)]

b21.3 = (σ2/ σ1 ) [(r12 – r23r13)/(1– r132)]

b23.1 = (σ2/ σ3 ) [(r23 – r12r13)/(1– r132)]

b31.2 = (σ3/ σ1 ) [(r13 – r23r12)/(1– r122)]

b32.1 = (σ3/ σ2 ) [(r23 – r13r12)/(1– r122)]

 35

Qn: If r12 = 0.7, r31 = r23 = 0.5, σ1 = 2, σ2 = 3 and σ3 = 3, find the equation of plane of regression x1
on x2 and x3.

Sol:

Here, means of the variables are not given, and therefore, it is convenient to write the equations
of planes of regression using Yule’s notation.

Equation of plane of regression x1 on x2 and x3 is :

X1 = b12.3X2+ b13.2X3

b12.3 = (σ1/ σ2 ) [(r12 – r13r23)/(1– r232)]

= ( 2/3) [(0.7 – 0.5 x 0.5)/(1– 0.52)]

= (2/3) [(0.7– 0.25)/(1– 0.25)]

= (2/3) [0.45/0.75] = (2/3) (0.6) = 0.4

b13.2 = (σ1/ σ3 ) [(r13 – r12r23)/(1– r232)]

= ( 2/3) [(0.5 – 0.7 x 0.5)/(1– 0.52)]

= (2/3) [(0.5– 0.35)/(1– 0.25)]

= (2/3) [0.15/0.75] = (2/3) (0.2) = 0.133

∴ X1 = 0.4X2+ 0.133X3

Qn: In a trivariate distribution, x̄1 =53, x̄2 = 52, x̄3 = 51, σ1 = 3.88, σ2 = 2.97, σ3 = 2.86, r23= 0.8, r31=
0.81 and r12= 0.78. Find the linear regression equation of x1 on x2 and x3.

Sol:

Here means of the variables are given.

∴ Regression equation of x1 on x2 and x3 is:

(x1 – x̄1) = b12.3(x2 – x̄2) + b13.2(x3 – x̄3)

b12.3 = (σ1/ σ2 ) [(r12 – r13r23)/(1– r232)]

= ( 3.88/2.97) [(0.78 – 0.81 x 0.8)/(1– 0.82)]

= (3.88/2.97) [(0.78– 0.648)/(1– 0.64)]

 36

= (3.88/2.97) [0.132/0.36] = (3.88/2.97) (0.367) = 0.4794

b13.2 = (σ1/ σ3 ) [(r13 – r12r23)/(1– r232)]

= ( 3.88/2.86) [(0.81 – 0.78 x 0.8)/(1– 0.82)]

= (3.88/2.86) [(0.81– 0.624)/(1– 0.64)]

= (3.88/2.86) [0.186/0.36] = 1.357 x 0.517 = 0.702

∴ (x1 – 53) = 0.4794(x2 – 52) + 0.702(x3 – 51)

= 0.4794 x2 – 24.929 + 0.702 x3 – 35.8

x1 = 0.4794 x2 + 0.702 x3 – 35.8– 24.929 + 53

= 0.4794 x2 + 0.702 x3 – 7.729

∴ x1 = 0.4794 x2 + 0.702 x3 – 7.729

x1 = 0.48x2 + 0.7x3 – 7.73

Qn: In a trivariate distribution, x̄1 =28.02, x̄2 = 4.91, x̄3 = 594, σ1 = 4.4, σ2 = 1.1, σ3 = 80, r23= –0.56,
r31= – 0.4 and r12= 0. 8. Estimate the value of x1 when x2 = 6 and x3 = 650.

Sol:

Here, to estimate the value of x1, we have to find the regression equation of x1 on x2 and x3.

Regression equation of x1 on x2 and x3 is:

(x1 – x̄1) = b12.3(x2 – x̄2) + b13.2(x3 – x̄3)

b12.3 = (σ1/ σ2 ) [(r12 – r13r23)/(1– r232)]

= ( 4.4/1.1) [(0. 8 – - 0.56 x -0.4)/(1– -0.562)]

= (4) [(0. 8– 0.224)/(1– 0.314)]

= 4 [0.576/0.6869] = 4 x 0.84 = 3.36

b13.2 = (σ1/ σ3 ) [(r13 – r12r23)/(1– r232)]

= ( 4.4/80) [(-0.4 – 0. 8 x -0.56)/(1– -0.562)]

= 0.055 [(-0.4 – - 0.448)/(1– 0.314)]

= 0.055 [(-0.4+ 0.448)/0.686] = 0.055 x [0.048/0.868]

= 0.055 x 0.07 = 0.00385

 37

∴ (x1 – 28.02) = 3.36(x2 – 4.91) + 0.00385(x3 – 594)

= 3.36 x2 – 16.498 + 0.00385 x3 – 2.287

x1 = 3.36x2 + 0.00385x3 – 16.498– 2.287 + 28.02

= 3.36 x2 + 0.00385x3 + 9.235

∴ x1 = 3.36x2 + 0.00385x3 + 9.235

REVIEW QUESTIONS:

1. What do you mean by regression analysis?

2. What are the different types of regression?
3. What do you mean by linear and non-linear regressions?
4. What do you mean by line of best fit?
5. What are the different methods for computing regression equations?
6. What do you mean by regression lines?
7. What are the important properties of regression coefficients?
8. What do you mean by multiple regressions?
9. Distinguish between correlation and regression analysis.
10. What do you mean by normal equation method for computing regression equations?
11. From the following data, fit the regression equations X on Y and Y on X:
X 102 80 100 88 84 82 90 96 97 83 79 88
Y 100 97 98 83 84 72 84 101 102 88 84 87
Also find the value of X, if Y 90 and Y if X = 105
12. In a trivariate distribution, x̄1 =10, x̄2 = 15, x̄3 = 12, σ1 = 3, σ2 = 4, σ3 = 5, r23= 0.4, r31= 0.6 and
r12= 0. 7. Determine the regression equation of X1 on X2 and X3.
 38

CHAPTER 4

PROBABILITY DISTRIBUTIONS
(THEORETICAL DISTRIBUTIONS)

Definition
Probability distribution (Theoretical Distribution) can be defined as a distribution obtained
for a random variable on the basis of a mathematical model. It is obtained not on the basis of
actual observation or experiments, but on the basis of probability law.
Random variable
Random variable is a variable who value is determined by the outcome of a random
experiment. Random variable is also called chance variable or stochastic variable.
For example, suppose we toss a coin. Obtaining of head in this random experiment is a random
variable. Here the random variable of “obtaining heads” can take the numerical values.
Now, we can prepare a table showing the values of the random variable and corresponding
probabilities. This is called probability distributions or theoretical distribution.
In the above, example probability distribution is :-
Obtaining of heads Probability of
(X) obtaining heads
P(X)
0 ½

1 ½
∑ P (X) = 1

Properties of Probability Distributions:

1. Every value of probability of random variable will be greater than or equal to zero.
i.e., P(X) ≤0
i.e., P(X) is always non-negative value

2. Sum of all the probability values will be 1

∑P(X) = 1
Question:
A distribution is given below. State whether this distribution is a probability distribution.

X: 0 1 2 3 4

P(X): 0.01 0.10 0.50 0.30 0.90

Solution
 39

Here all values of P(X) are more than zero; and sum of all P(X) value is equal to 1
Since two conditions, namely P(X) ≤0 and ∑P(X) = 1, are satisfied, the given distribution is a probability
distribution.
MATHEMATICAL EXPECTATION
(EXPECTED VALUE)
If X is a random variable assuming values x1, x2, x3,…………,xn with corresponding probabilities P1,
P2, P3,…………,Pn, then the Expectation of X is defined as x1p1+ x2p2+ x3p3+………+ xnpn.
E(X) = ∑ [x. p(x)]

Expected Value [ i. e, E (X) ] = ∑ [x. p(x)]

Qn:

A petrol pump proprietor sells on an average Rs. 80,000/- worth of petrol on rainy days and an
average of Rs. 95.000 on clear days. Statistics from the meteorological department show that the
probability is 0.76 for clear weather and 0.24 for rainy weather on coming Wednesday. Find the expected
value of petrol sale on coming Wednesday.

There are three alternative proposals before a business man to start a new project:-

Proposal I: Profit of Rs. 5 lakhs with a probability of 0.6 or a loss of Rs. 80,000 with a
probability of 0.4.

Proposal II: Profit of Rs. 10 laksh with a probability of 0.4 or a loss of Rs. 2 lakhs with a
probability of 0.6

Proposal III: Profit of Rs. 4.5 lakhs with a probability of 0.8 or a loss of Rs. 50,000 with a
probability of 0.2

If he wants to maximize profit and minimize the loss, which proposal he should prefer?
Sol:
Here, we should calculate the mathematical expectation of each proposal.
Expected Value E(X) = ∑ [x. p(x)]
Expected Value of Proposal I = (500000 x 0.6) + (80000 x 0.4) = 300000 – 32,000
= Rs. 2,68,000
Expected Value of Proposal II = (10,00.000 × 0.4) +(-2,00.000) = 400000 - 120000
= Rs: 2,80,000.
Expected Value of Proposal III = (450000 × 0.8) + ( - 50000 x 0.2) = 360000 - 10000
= Rs: 3,50,000
 40

Since expected value is highest in case of proposal III, the businessman should prefer the proposal III.

Classification of Probability Distribution

Following are the different types of probability distribution:

1. Binomial Distribution
2. Poisson Distribution
3. Uniform Distribution
4. Exponential Distribution
5. Normal Distribution

REVIEW QUESTIONS:
1. Define frequency distribution.
2. Define Random Variable.
3. What are the important properties of frequency distribution?
4. What is meant by Expected Value?
5. What are the different types of probability distributions?
 41

CHAPTER 5
BIONOMIAL DISTRIBUTION
Meaning & Definition:
Binomial Distribution is associated with James Bernoulli, a Swiss Mathematician.
Therefore, it is also called Bernoulli distribution. Binomial distribution is the probability
distribution expressing the probability of one set of dichotomous alternatives, i.e., success or
failure. In other words, it is used to determine the probability of success in experiments on which
there are only two mutually exclusive outcomes. Binomial distribution is discrete probability
distribution.
Binomial Distribution can be defined as follows: “A random variable r is said to follow
Binomial Distribution with parameters n and p if its probability function is:

P(r) = nC r prqn-r
Where, P = probability of success in a single trial
q=1–p
n = number of trials
r = number of success in ‘n’ trials.
Assumption of Binomial Distribution
(Situations where Binomial Distribution can be applied)
Binomial distribution can be applied when:-
1. The random experiment has two outcomes i.e., success and failure.
2. The probability of success in a single trial remains constant from trial to trial of the
experiment.
3. The experiment is repeated for finite number of times.
4. The trials are independent.
Properties (Features) of Binomial Distribution
1. It is a discrete probability distribution.
2. The shape and location of Binomial distribution changes as ‘p’ changes for a given ‘n’.
3. The mode of the Binomial distribution is equal to the value of ‘r’ which has the largest
probability.
4. Mean of the Binomial distribution increases as ‘n’ increases with ‘p’ remaining
constant.
5. The mean of Binomial distribution is np.
6. The Standard deviation of Binomial distribution is √npq
7. The variance of Binomial Distribution is npq
8. If ‘n’ is large and if neither ‘p’ nor ‘q’ is too close zero, Binomial distribution may be
approximated to Normal Distribution.
9. If two independent random variables follow Binomial distribution, their sum also
follows Binomial distribution.
 42

Qn: Six coins are tossed simultaneously. What is the probability of obtaining 4 heads?
Sol: P(r) = nC r prqn-r
r=4
n=6
p=½
q=1–p=1–½=½
฀ p ( r = 4) = 6C4 ( ½ )4 ( ½ )6-4

6! x (½)4+2
=
(6–4)!4!
6 ! x (½)6
=
2! 4!
6x5 1
= x
2x1 64

30
=
128
= 0.234
Qn: The probability that Sachin scores a century in a cricket match is 1/3. What is the probability that
out of 5 matches, he may score century in:
(1) Exactly 2 matches
(2) No match
Sol: Here p = 1/3 , n = 5, q = 2/3
P(r) = nC r prqn-r
(1) Probability that Sachin scores centuary in exactly 2 matches is:
P (r = 2) = 5C2 1/32 2/35-2
5! 1 8
= x x
(5–2)!2! 9 27
5X4 1 8
= x x
2X1 9 27

160
=
486
 43

80 = 0.329
=
243
(2) Probability that Sachin scores century in no match is:
P( r = 0) = 5C0 1/3 0 2/35-0
5! X 1 x 2/35
(5-0)! 0!

= 1 x 1 x (2/3)5

32
=

243

= 0.132
Qn: Consider families with 4 children each. What percentage of families would you expect to have :-
(a) Two boys and two girls
(b) At least one boy
(c) No girls
(d) At the most two girls
(a) P( having a boy) = ½
P (having a girl) = ½
n = 4
P (getting 2 boys & 2 girls) = p (getting 2 boys)
= p (r = 2) = 4C2 (½) 2 (1/2)4-2

= 4! x (1/2)2 x (½)2
(4-2)! 2!

= 4 x 3 x (1/2)4
2
= 6 x 1/16 = 6/16 = 3/8

∴ Percentage of families with 2 boys and 2 girls = (3/8) x 100 = 37.5 % .

(b) Probability of having at least one boy:

= p (having one boy or having 2 boys or having 3 boys or having 4 boys)
= p (having one boy) + p (having 2 boys) + p (having 3 boys)
+ p (having 4 boys)
= p (r=1) + p (r = 2) + p(r = 3) + p (r = 4)

= 4/16 + 6/16 + 4/46 + 1/16 = 15/16

∴ Percentage of families with at least one boy = (15/16) x 100 = 93.75 %

 44

(c) Probability of having no girls = Probability of having 4 boys

P (r = 4) = 4C4 (½) 4 (½) 4-4 = 1 x (½ )4 = 1/16
∴ Percentage of families with at least one boy = (1/16) x 100 = 6.25 %
(d) Probability of having at the most 2 girls = P(having 2 or 1 or 0 girls)
= P(having 2 boys or 3 boys or 4 boys)
= 11/16.
∴ Percentage of families with at least one boy = (11/16) x 100 = 68.75 %

Qn: For a binomial distribution mean = 4 and variance = 12/9. Find n.

Sol. Mean np = 4 ……………….. (1)

Variance npq = 12/9 …………… (2)

Divide (2) by (1):

We get q = 12/9 ÷ 4 = 12/36 = 1/3

∴p = 1 - 1/3 = 2/3

∴ n x 2/3 = 4, n = 4 x 3/2 = 6

n = 6

Fitting a Binomial Distribution

Steps:
1. Find the value of n, p and q
2. Substitute the values of n, p and q in the Binomial Distribution function of nC r prqn-r
3. Put r = 0, 1, 2, ............... in the function nC r prqn-r
4. Multiply each of such terms by total frequency (N) to obtain the expected frequency.

Qn: Eight coins were tossed together for 256 times. Fit a Binomial Distribution of getting heads.
Also find mean and standard deviation.

Sol: p (getting head in a toss) = ½ , n = 8, q = ½

Binomial Distribution function is p(r) = nC r prqn-r

Put r = 0, 1, 2, 3 .............. 8, then are get the terms of the Binomial Distribution.

Binomial Distribution

No. of Heads (x) P(x) Expected Frequency =

P(x) x 256

0 8C0 (1/2)0 (1/2) 8 = 1/256 1

8C1 (1/2)1 (1/2) 7 = 8/256
1 8
2 6
8C2 (1/2) (1/2) = 28/256
2 28
 45

8C3 (1/2)3 (1/2) 5 = 56/256

3 56
4 4
8C4 (1/2) (1/2) = 70/256
4 70
8C5 (1/2)5 (1/2) 3 = 56/256
5 53
8C6 (1/2)6 (1/2) 2 = 28/256
6 28
7 1
8C7 (1/2) (1/2) = 8/256
7 8
8C8 (1/2)8 (1/2) 0 = 1/256
8 1
Total
256

Mean = np = 8*1/2 = 4

S.D = √npq = √8*1/2* ½ = √2 = 1.414

REVIEW QUESTIONS:
1. Define Binomial Distribution.
2. What are the important properties of Binomial Distribution?
3. Examine whether the following statement is true:
“ For a Binomial Distribution, mean = 10 and S D = 4”
4. For a Binomial Distribution, mean = 6 and S D = √2. Find parameters. Write down all the
terms of the distribution.
*********
 46

CHAPTER 6

POISSON DISTRIBUTION
Meaning and Definition
Poisson distribution is a limiting form of Binomial Distribution. In Binomial distribution,
the total number of trials is known previously. But in certain real life situations, it may be
impossible to count the total number of times a particular event occurs or does not occur. In such
cases Poisson distribution is more suitable.
Poison Distribution is a discrete probability distribution. It was originated by Simeon
Denis Poisson.
A random variable “r” said to follow Binomial distribution if its probability function is:

e –m . mr
P ( r) = r!

Where r = random variable (i.e., number of success in ‘n’ trials)

e = 2.7183
m = mean of Poisson distribution.
Properties of Poisson Distribution
1. Poisson distribution is a discrete probability distribution.
2. Poisson distribution has a single parameter ‘m’. When ‘m’ is known all the terms can
be found out.
3. It is a positively skewed distribution.
4. Mean and Variance of Poisson distribution are equal to ‘m’.
5. In Poisson distribution, the number of success is relatively small.
6. Standard deviation of Poisson distribution is √m.
Practical situations where Poisson distribution can be used
1. To count the number of telephone calls arising at a telephone switch board in a unit of
time.
2. To count the number of customers arising at the super market in a unit of time.
3. To count the number of defects in Statistical Quality Control.
4. To count the number of bacteria per unit.
5. To count the number of defectives in a park of manufactured goods.
 47

6. To count the number of persons dying due to heart attack in a year.

7. To count the number of accidents taking place in a day on a busy road.

Qn: A fruit seller, from his past experience, knows that 3 of apples in each basket will be defectives.
What is the probability that exactly 4 apples will be defective in a given basket?
Sol. m = 0.03

e –m . mr
P ( r) = r!

∴ P ( exactly 4 apples are defective) = (e-3 . 34) / 4!

= (0.0498 x 81) /24

= 0.16807

Qn: It is known from the past experience that in a certain plant, there are on an average four
industrial accidents per year. Find the probability that in a given year there will be less
than four accidents. Assume Poisson distribution.
Sol:

e –m . mr
P ( r) = r!

m=4
∴ P ( exactly 4 apples are defective) = P (r < 4)
P (r < 4) = P (r = 0 or 1 or 2 or 3)
= P (r = 0) + P (r =1) + P (r = 2) + P (r = 3)
P (r = 0) = (e-4 . 4 0) / 0! = (0.0183 x 1) / 1 = 0.0183
P (r = 1) = (e-4 . 4 1) / 1! = (0.0183 x 4) / 1 = 0.0732
P (r = 2) = (e-4 . 4 2) / 2! = (0.0183 x 16) / 2 = 0.1464
P (r = 3) = (e-4 . 4 3) / 3! = (0.0183 x 64) / 6 = 0.1952
∴ P (r < 4) = 0.0183+ 0.0732 + 0.1464 + 0.1952 = 0.4331
Qn: Out of 500 items selected for inspection, 0.2% is found to be defective. Find how many lots
 48

will contain exactly no defective if there are 1000 lots.

Sol:

e –m . mr
P ( r) = r!

m = 500 x 0.2% = 1

∴ P (r = 0) = (e -1 10) / 0! = ( 0.3679 x 1) / 1 = 0.3679

∴ No. of lots having zero defective = 0.3679 x 1000 = 368

Qn: In a certain factory producing optical lenses, there is a small chance of 1/500 for any one lens to
be defective. The lenses are supplied in packets of 10. Use P.D to calculate the approximate number of
packets containing no defectives, one defective, two defectives and three defective lenses respectively
in a consignment of 20,000 packets.

Sol:

e –m . mr
P ( r) = r!

m = 10 x 1/500 = 0.02

∴ P (r = 0) = (e-0.02 x 0.02 0 ) / 0! = (0.9802 x 1) /1 = 0.9802

∴ No. of packets containing no defective lens = 0.9802 x 20000 = 19604

P (r = 1) = (e-0.02 x 0.02 1 ) / 1! = (0.9802 x 0.02) /1 = 0.0196

∴ No. of packets containing no defective lens = 0.0196 x 20000 = 392

P (r = 2) = (e-0.02 x 0.02 2 ) / 2! = (0.9802 x 0.0004) /2 = 0.00019604

∴ No. of packets containing no defective lens = 0.00019604 x 20000 = 4

P (r = 3) = (e-0.02 x 0.02 3 ) / 3! = (0.9802 x 0.000008) /6 = 0.0000013069

∴ No. of packets containing no defective lens = 0.0000013069 x 20000 = 0

Qn: A Systematic sample of 100 pages was taken from a dictionary and the observed frequency
distribution of foreign words per page was found to be as follows:
No. of foreign words per page (x) : 0 1 2 3 4 5 6
 49

Frequency (f) : 48 27 12 7 4 1 1
Calculate the expected frequencies using Poisson distribution.
Sol: At first, we have to know the parameter of P.D, which is equal to the mean of the given
distribution. So find the mean of the distribution:
Mean = (εfx) / εf

x 0 1 2 3 4 5 6

f 48 27 12 7 4 1 1 N = εf = 100

fx 0 27 24 21 16 5 6 (εfx) = 99
Mean = 99/100 = 0.99

Calculation of Expected Frequencies

X P(x) = (e-m . mx) / x! Expected Frequency = P(x) . N
0 (e-0.99 . 0.990)/0! =(0.3716 x1)/1 = 0.3716 0.3716 x 100 = 37.16 = 37
1 (e-0.99 . 0.991)1! =(0.3716 x 0.99)/1= 0.3679 0.3716 x 100 = 37.16 = 37
2 (e-0.99 . 0.992)2! =(0.3716 x 0.98)/2= 0.1821 0.1821 x 100 = 18.21 = 18
3 (e-0.99 . 0.993)3! =(0.3716 x 0.97)/6= 0.0601 0.0601 x 100 = 6.01 = 6
4 (e-0.99 . 0.994)4!=(0.3716 x 0.96)24= 0.0149 0.0149 x 100 = 1.49 = 2
5 (e-0.99 . 0.995)5!=(0.3716 x 0.95)120= 0.0029 0.0029 x 100 = 0.29 = 0
6 (e-0.99 . 0.996)6!=(0.3716 x 0.94)720= 0.0005 0.0005 x 100 = 0.05 = 0
Total 100

REVIEW QUESTIONS:
1. Define Poisson distribution.
2. What are the important properties of P.D?
3. What are the situations under which P D can be applied?
4. Write down the probability function of P.D. whose mean is 2. What is its variance?
5. A machine is producing 4% defectives. What is the probability of getting at least 4 defectives
in a sample of 50 =, using (a) BD and (b) PD?
6. The following table gives the number of days in a 50 day period during which automobile
accidents occurred in a certain part of the city. Fit a Poisson distribution to the data:
No. of accidents 0 1 2 3 4
No. of days 19 18 8 4 1
********
 50

CHAPTER 7

NORMAL DISTRIBUTION
Meaning and Definition
The normal distribution is a continuous probability distribution. It was first developed by
De-Moivre in 1733 as limiting form of binomial distribution. Fundamental importance of normal
distribution is that many populations seem to follow approximately a pattern of distribution as
described by normal distribution. Numerous phenomena such as the age distribution of any
species, height of adult persons, intelligent test scores of students, etc. are considered to be
normally distributed.

Definition of Normal Distribution

A continuous random variable, ‘X’, said to follow Normal Distribution if its probability
function is:

2
1 -- ½ (x-μ)/σ
P (x) = e

√2π . σ
 51

Properties of Normal Distribution (Normal Curve)

1. Normal distribution is a continuous distribution.

2. Normal curve is symmetrical about the mean.
3. Both sides of normal curve coincide exactly.
4. Normal curve is a bell shaped curve.
5. Mean, Median and Mode coincide at the centre of the curve.
6. Quantities are equi-distant from median. Q3 – Q2 = Q2 – Q1
7. Normal curve is asymptotic to the base line.
8. Total area under a normal curve is 100%.
9. The ordinate at the mean divide the whole area under a normal curve into two equal parts.
(50% on either side).
10. The height of normal curve is at its maximum at the mean.
11. The normal curve is unimodel, i.e., it has only one mode.
12. Normal curve is mesokurtic.
13. No portion of normal curve lies below the x-axis.
14. Theoretically, the range of normal curve is – ∞ to + ∞ . But practically the range is μ -
3σ to μ + 3σ.
15. Area under the normal curve is distributed as follows: (Area property)
(a) μ ± σ covers 68.27% area
(b) μ ± σ covers 95.45% area
(c) μ ± σ covers 99.73% area

Importance or Uses of Normal Distribution

The normal distribution is of central importance because of the following reasons:
1. The discrete probability distributions such as Binomial distribution and Poisson
distribution tend to normal distribution as ‘n’ becomes large.
2. Almost all sampling distributions conform to the normal distribution for large values of
‘n’.
3. Many tests of significance are based on the assumption that the parent population from
which samples are drawn follows normal distribution.
4. The normal distribution has numerous mathematical properties which make it popular and
comparatively easy to manipulate.
5. Normal distribution finds applications in Statistical Quality Control.
6. Many distributions in social and economic data are approximately normal. For example,
birth, death, etc. are normally distributed.

Area under Standard Normal Curve

 52

In case of normal distribution, probability is determined on the basis of area. But

the area we have to calculate the ordinate of z – scale.

The scale to which the standard deviation is attached is called z-scale.

Z = (x – μ) / σ

Qn: The variable, x, follows normal distribution with mean = 45 and S.D = 10. Find the
probability that x ≥ 60.

Sol: μ = 45, σ = 10, x = 60

Z = (x – μ) / σ

Z = (60 – 45) / 10 = 15/10 = 1.5

0.4332

P (x ≥ 60) means P ( z ≥ 1.5)

= 0.5 - 0.4332 = 0.0668

P (x ≥ 60) = 0.0668

Qn: The variable, x, follows normal distribution with mean = 45 and S.D = 10. Find the
probability that x ≤ 40.

Sol: μ = 45, σ = 10, x = 40

Z = (x – μ) / σ

Z = (40 – 45) / 10 = -5/10 = -- 0.5

0.1915

P (x ≤ 40) means P ( z ≥ - 0.5)

= 0.5 - 0.1915 = 0.3085

 53

P (x ≤ 40) = 0.3085

Qn: The variable, x, follows normal distribution with mean = 45 and S.D = 10. Find the
probability that 40 ≤ x ≤ 56.

Sol: μ = 45, σ = 10, x1 = 40, x2 = 56

Z = (x – μ) / σ
When x = 40, Z = (40 -- 45) / 10 = -5 / 10 = - 0.5
When x = 56, Z = (56 -- 45) / 10 = 11 / 10 = 1.1

0.1915 0.3643

P (40 ≤ x ≤ 56) means P ( z – 0.5 ≤ x ≤ 1.5)

= 0.1915 + 0.3643 = 0.5558

P (40 ≤ x ≤ 56) = 0.5558

Qn: The scores of students in a test follow normal distribution with mean = 80 and S D = 15. A
sample of 1000 students has been drawn from the population. Find (1) probability that a randomly
chosen student has score between 85 and 95 (2) appropriate number of students scoring less than 60.

Sol.
(1) μ = 80, σ = 15, x1 = 85, x2 = 95

Z = (x – μ) / σ

When x = 85, Z = (85 – 80) / 15 = 5/15 = 0.333

When x = 95, Z = (95 – 80) / 15 = 15/15 = 1

0.1293 0.212
 54

∴ P (85 ≤ x ≤ 95) = P (0.333 ≤ z ≤ 1) = 0.3413 – 0.1293 = 0.212

Probability that a student scores between 85 and 95 = 0.212
(2) P (Less than 60):
When x = 60,
Z = (x – μ)/σ = (60 -- 80)/15 = -20/15 = - 1. 333

0.0918 0.4082

P(x < 60) = P(z < -1.333) = 0.5 -- 0.4082 = 0.0918

∴ Number of students scoring less than 60 = 0.0918 x 1000 = 91.8 = 92 students

Computation of Z-value when Area is known

Qn: In a competitive examination, 5000 candidates have appeared. Their average mark was
62 and S.D was 12. If there are only 100 vacancies, find the minimum marks that one should
score in order to get selection.

Sol: 𝛍 = 62, σ = 12
Number of vacancies = 100
Percentage of vacancies to the total number of candidates = (100/5000) x 100 = 2% = 0.02
Area corresponds to the students who will get selection is shown in the following normal
curve:

0.02

Therefore, the area to the left of the above area of 0.02 is:

Z = (40 – 45) / 10 = -5/10 = -- 0.5

0.48
 55

Locate the area of 0.48 in the table and find the Z – value corresponds to it.
The table shows the area nearest to 0.48 is 0.4798, and the corresponding z-value is 2.05
Z = 2.05
(x – μ)/σ = 2.05
(x -- 62)/12 = 2.05, x – 62 = 2.05 x 12
x -- 62 = 24.6, ∴ x = 24.6 + 62 = 86.6
∴ The minimum marks one should score to get section = 86.6 marks

Construction of Normal Distribution

Procedure:

1. Find the mean and S.D of the given distribution and take them as μ and σ
(parameters) of the normal distribution.
2. Take the lower limit of each class as the x values.
3. Calculate the z-value corresponding to each x-value by using formulae z = (x—μ)/σ.
Z-value of first and last values need not be computed.
4. Find the area corresponds to z-value from the standard normal distribution table. The
area corresponds to the first and last z-values will be 0.5.
5. Find the area of each class using the area (probability) of respective class limits.
(Take the difference in case of same signs; and take the total in case of opposite signs)
6. Multiply tye area of ech class by the total frequency to the frequency of the class.
The new frequency distribution with theoretical frequencies will be a normal
approximation to the given frequency distribution.
Qn: Fit a normal distribution to the following data:

X 10-20 20-30 30-40 40-50 50-60 60-70 70-80

f 4 22 48 66 40 16 4
Sol:

Computation of Mean and Standard deviation

Class Mid F d d’ fd’ d’2 fd’2
point (m) (m-35)
10-20 15 4 -20 -2 -8 4 16
20-30 25 22 -10 -1 -22 1 22
30-40 35 48 0 0 0 0 0
40-50 45 66 10 1 66 1 66
50-60 55 40 20 2 80 4 160
60-70 65 16 30 3 48 9 144
70-80 75 4 40 4 16 16 64
200 180 472

x̄ = A + [(εfd’)/N] x C, x̄ = 35 + [(180/200)x10], = 35+9 = 44

2 2
S. D = √(εfd’ /N) - [(εfd’)/N] x 10 = √1.55 x 10 = 12.45
 56

∴ μ = 44 and σ = 12.45

Computation of Expected Frequencies

Lower Z= (x—μ)/σ Area Area of class Expected
limit Frequency
10 0.5000
-2.73
0.0268 5
20 0.4732
-1.93
0.1046 21
30 0.3686
-1.12
0.2431 49
40 0.1255
-0.32
0.3099 62
50 0.48 0.1884
0.2171 43
60 1.29 0.4015
0.0802 16
70 2.09 0.4817
0.0183 4
80 2.89 0.5000
Total 200

Review Questions:
1. Define normal distribution.
2. What are the important properties of normal distribution?
3. Explain the importance of normal distribution.
4. Explain the procedure for construction of normal distribution.
5. If x follows a normal distribution with mean 12 and variance 16, find P(x≥20).
6. The weekly wages of 1000 workers are normally distributed with mean of 70 and S.D
of 5. Estimate the number of workers whose wages lie between 69 and 72.
7. In an aptitude test administered to 900 students, the mean score is 50 and S.D is 20.
Find the number of students securing scores (a) between 30 and 70 (b) exceeding 65.
Find the value of the score exceeded by the top 90 students.
8. Construct a normal distribution to the following data of marks obtained by 100
students:
Marks 60-62 63-65 66-68 69-71 72-74
No. of Students 5 18 42 27 8

*********
 57

Chapter 8

EXPONENTIAL DISTRIBUTION
Definition of Exponential Distribution

A continuous random variable, x, follows random variable if its probability density

function is :

f (x) =  e -x for x > 0 and  > 0

Parameter of Exponential Distribution

The single parameter of exponential distribution is . If we know , we can find out
all the terms.

Properties of Exponential Distribution

1. The mean of Exponential distribution is = 1/
2. The variance of Exponential distribution is = 1/2
3. The first four moments of Exponential distribution are:
μ 1 = 1/ , μ 2 = 1/2 , μ 3 = 2/3 , μ 4 = 9/4
4. The measure of Skewness of Exponential distribution is = √β1 = 2
5. The measure of Kurtosis of Exponential distribution is = β2 = 9
6. The Median of Exponential distribution is = 1/

REVIEW QUESTIONS:

1. Define exponential distribution.

2. Write down the first four moments of exponential distribution.
3. What is the skewness of exponential distribution?
4. What about the median of an exponential distribution?
5. What are the important properties of exponential distribution?
 58

Chapter 9

UNIFORM DISTRIBUTION
Definition of Uniform Distribution

A discrete random variable, x, follows uniform distribution if its probability density function
is :

f (x) = 1/n for x = x1, x2, x3, .........................., xn

For example, when a die is thrown, let x stands for the numbers obtained.

Then f (x) = 1/6 for x = 1, 2, 3, 4, 5, 6.

Mean of Uniform Distribution

Mean of Uniform Distribution = ε x/n

Variance of Uniform Distribution

Variance of Uniform Distribution = [ε x2/n] – [ε x/n]2

REVIEW QUESTIONS:

1. Define Uniform distribution.

2. What about the mean of uniform distribution?
3. What about the variance of uniform distribution?
 59

CHAPTER 10
STATISTICAL INFERENCE
Basic Concepts
Population: In statistics, ‘Population’ refers to collection of all individuals or objects or
items or things under consideration.
Finite Population: If a population contains a finite number of objects, it is called finite
population. Eg: Students in a college.
Infinite Population: If a population contains a infinite number of objects, it is called infinite
population. Eg: Stars in the sky.
Sample: A sample is a representative part of the population.
Sample size: Number of units in a sample group is called sample size. If sample size is too
small, it may not represent the population. If it is very large, it may require more time and
money for investigation. Hence, the size of a sample should be optimum.
Large Sample: If the size of a sample exceeds thirty, it is called as large sample.
Small Sample: If the size of a sample does not exceed thirty, it is called as small sample.
Parameter: It is a statistical measure derived from population elements. If the arithmetic
mean is computed from all the elements of a population, it is a population parameter. Here it
is called population mean. Population mean is denoted by the symbol μ. Population standard
deviation is denoted by σ.
Statistic: It is a statistical measure derived from sample elements. If the arithmetic mean is
computed from the elements of a sample group, it is a sample statistic. Here it s called sample
mean. Sample mean is denoted by the symbol x̄. Sample standard deviation is denoted by ‘s’.

Statistical inference
Statistical inference refers to the process of selecting samples and using sample statistic to
draw inference or conclusion about the population parameter or population distribution. The
two main branches of statistical inference are:
(a) Testing of Hypothesis
(b) Estimation
Testing of Hypothesis
Testing of hypothesis is the process under which a statistical hypothesis about a population is
formulated and its validity is tested on the basis of a random sample drawn from that
population. For testing the validity of a hypothesis, a number of tests are used. All these tests
can be classified into two categories, namely (i) parametric tests and (ii) non-parametric
tests. Z-test, t-test, Chi-square test, F-test, etc. are commonly used statistical tests.

Procedure for testing hypothesis

(1) Set up null hypothesis and alternative hypothesis
(2) Decide the test statistic (statistical test) applicable for testing hypothesis.
(3) Apply the appropriate formulae for computing the value of the test statistic.
(4) Specify the level of significance. If nothing is mentioned about the level of
significance, take 5% level o significance.
(5) Fix the degree of freedom
 60

(6) Locate the table value (critical value) of the test statistic at specified level of
significance
(7) Compare the calculated value of the test statistic with the corresponding table value
(critical value) and decide whether to accept or reject the null hypothesis. If calculated
value of the test statistic is numerically less than the table value, the null hypothesis is
accepted. If calculated value of the test statistic is numerically more than the table
value, the null hypothesis is rejected.

Hypothesis
Hypothesis is a tentative solution or assumption or proposition about the parameter or
nature of the population. It is a logically drawn conclusion about the population.
Null Hypothesis
This is the original hypothesis. A null hypothesis is a hypothesis which formulated for
the purpose of rejection. The term “null” refers to ‘nil’ or ‘no’ or ‘amounting to
nothing’. This hypothesis is generally set up as there is no significant difference
between the sample statistic and population parameter. A null hypothesis is denoted
by H0
Alternative Hypothesis
Any hypothesis other than null hypothesis is called alternative hypothesis. It is the
hypothesis which is accepted when the null hypothesis is rejected. An alternative
hypothesis is denoted by H1 or Ha
Sampling Distribution
Sampling distribution is a distribution of sample statistic derived from various samples
drawn from the same population. Since sample statistic is a random variable,
sampling distribution is a probability distribution.
Standard Error
Standard Error (SE) of a statistic is the standard deviation of the sampling distribution
of that statistic. For example, the Standard deviation of the sampling distribution of
the sample mean is σ/√n, where σ = population S.D. and n = sample size. Therefore
the Standard Error (SE) of sampling distribution of mean is σ/√n .
Uses of Standard Error
(1) Standard Error is used for testing a given hypothesis.
(2) Standard Error gives an idea about the reliability of a sample. The reciprocal of
Standard Error is a measure of reliability of the sample.
(3) Standard Error can be used to determine the confidence limits for population
values like mean, proportion and standard deviation.
Errors in Testing of Hypotheses
In any test of hypothesis is the decision is to accept or to reject a null hypothesis. The
 61

decision is based on the information supplied by the sample data. The four
possibilities of the decision are:
(1) Accepting a null hypothesis when it is true
(2) Rejecting a null hypothesis when it is false
(3) Rejecting a null hypothesis when it is true
(4) Accepting a null hypothesis when it is false
It is clear that the possibilities (1) and (2) are correct decisions. But the
possibilities (3) and (4) are errors.
Type I Error:
The error which is committed by rejecting the null hypothesis even when it is
true is called Type I error. It is denoted by alpha (α).
Type II Error:
The error which is committed by accepting the null hypothesis even when it is
wrong is called Type II error. It is denoted by beta (β).
When we try to reduce the possibility for one error, the possibility for the
other will be increased. Therefore, a compromise of these two is to be ensured. Type
II error is more dangerous than Type I error.
Power of a Test
Probability for rejecting the null hypothesis when the alternative hypothesis is true is
called power of a test.
Power of a test = 1 – P(Type II Error)
Level of Confidence
Level of confidence is the probability of accepting a true null hypothesis.
Level of Confidence = 1 – Level of significance.
If Level of significance is 5%, Level of Confidence = 95%.
Level of Significance
Level of Significance is the probability of rejecting a true null hypothesis. Level of
Significance is denoted by alpha (α). If nothing is mentioned about the level of
significance, it is taken as 5%.
Level of Significance (α) = 1 – level of acceptance
Acceptance Region
The area under the normal curve which represents the acceptance of a null hypothesis
(i.e; level of confidence) is called the Acceptance Region or Acceptance Area.
Acceptance Region = 100% -- Rejection Region
Rejection Region (Critical Region)
The area under the normal curve which represents the rejection of a null hypothesis
(i.e; level of significance) is called the Rejection Region or Critical Region.
Rejection Region = 100% -- Acceptance region
Degree of Freedom
Degree of freedom is defines as the number of independent observations which is
obtained by subtracting the number of constraints from the total number of
observations.
Degree of freedom (d.f) = Total No. observations – No. of constraints.
 62

Two-tailed Test
A two tailed test is one in which we reject the null hypothesis if the computed value
of the test statistic is significantly greater than or lower than the critical value (table
value) of the test statistic. Thus in two tailed tests the critical region is represented by
both tails. If we test the hypothesis at 10% level of significance, the size of the
acceptance region is 90% and the size of the rejection region is 10% on both sises
together.

45% 45%

5% Acceptance region 5%

Rejection Region

One-tailed Test
One tailed test is one in which the rejection region is located in only one tail of the
normal curve. It may be at left tail or right tail, depending on the alternative
hypothesis. If the alternative hypothesis is with ‘<’ (less than) sign, the rejection
region is placed on the left tail, and the test is called left-tailed test. If the alternative
hypothesis is with ‘>’ (more than) sign, the rejection region is placed on the right tail,
and the test is called right-tailed test.

45% 50%

5% Acceptance region

Rejection Region
(Left-tailed Test)

50% 45%

Acceptance region 5%

Rejection Region
(Right-tailed Test)
 63

Critical Value (Table Value)

The critical value is the value of the test statistic which separates the rejection region
from the acceptance region. It depends on the level of significance and degree of
freedom. When the calculated value of the test statistic is numerically less than the
critical value, the null hypothesis is accepted. When the calculated value of the test
statistic is numerically more than the critical value, the null hypothesis is accepted.

Parametric Tests
When testing of hypothesis is done, if some assumptions are made about the nature of
population distribution, then the test statistic applied there is called parametric test.
There are number of parametric tests. Eg: t-test, Z test, F test, etc.

Non-Parametric Tests
When testing of hypothesis is done, if no assumptions are made about the nature of
population distribution, then the test statistic applied there is called non-parametric
test. There are number of non-parametric tests. Eg: Chi-square Test, Sign tests, Signed
Rank Tests, Rank Sum Tests, Run Test, Kolmogrov Smirnov Test, etc. Since, no
assumptions are made about the nature of population, non-parametric tests are also
called distribution-free tests.

TESTING OF GIVEN POPULATION MEAN

This testing of hypothesis is used to test whether the given population mean is true or
not. In other words, it is used to test whether there is significant difference between
sample mean and population mean.
Procedure:
1. Set up H0 and H1
H0 : There is no significant difference between sample mean and population mean
( i.e; μ = μ0)
H1 : There is significant difference between sample mean and population mean
( i.e; μ ≠ μ0)
2. Decide the test statistic:
The test statistic applicable here is Z-test or t-test.
If population S.D.(i.e; σ) is known, apply Z-test
If population S.D.(i.e; σ) is unknown but sample is large, apply Z-test
If population S.D.(i.e; σ) is unknown but sample is small, apply t-test
3. Apply the appropriate formula for computing the value of the test statistic:

Z / t = Difference/Standard Error
Difference = Difference between sample mean and the given population mean
Standard Error = σ / √n ( If population S.D is known)
Standard Error = s / √n ( If population S.D is unknown, but sample is large)
 64

Standard Error = s / √n-1 ( If population S.D is unknown and sample is small)

Where σ = population S.D, s = sample S.D, n = sample size
4. Specify the level of significance. If nothing is mentioned about the level of
significance, take 5%.
5. Fix the degree of freedom:
For Z-test, d.f = infinity; For t-test, d.f = n-1
6. Locate the table value (critical value) of the test statistic at specified level of
significance and fixed degree of freedom.
7. Compare the calculated value of test statistic with the table value and decide
whether to accept or reject the null hypothesis. If calculated value of the test
statistic is numerically less than the table value, the null hypothesis is accepted. If
calculated value of the test statistic is numerically more than the table value, the
null hypothesis is rejected.

Qn: The mean life of random sample of 100 tyres is 15269 km. The manufacturer
claims that the average life of tyres manufactured by the company is 15200 km with
SD of 1248 km. Test the validity of company’s claim.

Sol:
H0 : There is no significant difference between sample mean and population mean
( i.e; μ = 15200)
H1 : There is significant difference between sample mean and population mean
( i.e; μ ≠ 15200)
Since population S.D is known, the test statistic applicable here is Z-test
Z = D/SE
D = x̄ - μ = 15269-15200 = 69
S E = σ/√n = 1248/√100 = 1248/10 = 124.8
Z = 69/124.8 = 0.553
Level of significance = 5%
Degree of freedom = infinity (population S D is known)
Table value (Critical value) at 5 % level of significance and infinity degree of
freedom is 1.96
Since calculated value of Z is less than the critical value, H0 is accepted. That is, there
is no significant difference between sample mean and population mean. μ =
15200. So, we may conclude that the claim of the company is valid.

Qn: A sample of size 400 was drawn and the sample mean was found to be 99. Test
whether this sample could have come from the normal population with mean = 100 ad
S.D = 8 at 5% level of significance.
 65

Sol:
H0 : There is no significant difference between sample mean and population mean
( i.e; μ = 100)
H1 : There is significant difference between sample mean and population mean
( i.e; μ ≠ 100)
Since population S.D is known, the test statistic applicable here is Z-test
Z = D/SE
D = x̄ - μ = 100 -- 99 = 1
S E = σ/√n = 8/√400 = 8/20 = 0.4
Z = 1/0.4 = 2.5
Level of significance = 5%
Degree of freedom = infinity (population S D is known)
Table value (Critical value) at 5 % level of significance and infinity degree of
freedom is 1.96
Since calculated value of Z is more than the critical value, H0 is rejected. H1 is
accepted. That is, there is significant difference between sample mean and
population mean. So, we may conclude that μ ≠ 100

Qn: A random sample of 200 bottles of talcum powder gave an average weight of
49.5 gram with a S.D of 2.1 gram. Do we accept the hypothesis of weight per
bottle is 50 gram at 1% level of significance?
Sol:
H0 : There is no significant difference between sample mean and population mean
( i.e; μ = 50)
H1 : There is significant difference between sample mean and population mean
( i.e; μ ≠ 50)
Since sample is large, the test statistic applicable here is Z-test
Z = D/SE
D = x̄ - μ = 50 – 49.5 = 0.5
S E = s/√n = 2.1/√200 = 2.1/14.142 = 0.148
Z = 0.5/0.148 = 3.378 (Calculated value)
Level of significance = 1%
Degree of freedom = infinity (population is large)
Table value (Critical value) at 1 % level of significance and infinity degree of
freedom is 2.58
Since calculated value of Z is more than the critical value, H0 is rejected. H1 is
accepted. That is, there is significant difference between sample mean and
 66

population mean. So, we may conclude that μ ≠ 50 gram

Qn: The average life of 26 bulbs were found to be 1200 hours with a S.D of 150 hours. Test
whether these bubs could be considered as a random sample from a normal population with
mean 1300 hours.
Sol: H0 : There is no significant difference between sample mean and population mean
( i.e; μ = 1300)
H1 : There is significant difference between sample mean and population mean
( i.e; μ ≠ 1300)
Since sample is small, the test statistic applicable here is t-test
t = D/SE
D = x̄ - μ = 1300 – 1200 = 100
S E = s/√n-1 = 150/√26-1 = 150/5 = 30
t = 100/30 = 3.333 (Calculated value)
Level of significance = 5%
Degree of freedom = 26-1 = 25 (sample is small)
Table value (Critical value) at 5% level of significance and 25 degree of freedom is
2.06
Since calculated value of Z is more than the critical value, H0 is rejected. H1 is
accepted. That is, there is significant difference between sample mean and
population mean. So, we may conclude that the bulbs could not e drawn from the
normal population with mean 1300 hours ( i.e; μ ≠ 1300).

Qn: A typist claims that he can type at a speed of more than 120 words per minute. Of the 12
tests given to him, he could perform an average of 135 words with a S.D of 40. Is his claim
valid at 1% level of significance?
Sol: H0 : There is no significant difference between sample mean and population mean
( i.e; μ = 120)
H1 : There is significant difference between sample mean and population mean
( i.e; μ > 120)
Here, the test One-tailed test ( Right tailed test)
Since sample is small, the test statistic applicable here is t-test
t = D/SE
D = x̄ - μ = 135 – 120 = 15
S E = s/√n-1 = 40/√12-1 = 40/√11 = 40/3.32 = 12.05
t = 15/12.05 = 1.245 (Calculated value)
Level of significance = 1%
 67

Degree of freedom = 12-1 = 11 (sample is small)

Table value (Critical value) at 1% level of significance and 11 degree of freedom is
2.718
Since calculated value of t is less than the critical value, H0 is accepted. μ = 120. That
is, there is no significant difference between sample mean and population mean.
So, we may conclude that the claim of the typist that he can type at a speed of
more than 120 words is not valid.

TESTING OF SIGNIFICANCE OF THE DIFFERENCE BETWEEN TWO SAMPLE

MEANS

This testing of hypothesis is used to test whether the difference between two sample means
are significant or not. If the difference is not significant, they are treated as equal; or we may
think that the two samples are drawn from the same population.
Procedure:
1. Set up H0 and H1
H0 : There is no significant difference between two sample means( i.e; μ1 = μ2)
H1 : is no significant difference between two sample means( i.e; μ1 ≠ μ2)
2. Decide the test statistic:
The test statistic applicable here is Z-test or t-test.
If population S.D.(i.e; σ) is known, apply Z-test
If population S.D.(i.e; σ) is unknown but sample is large, apply Z-test
If population S.D.(i.e; σ) is unknown but sample is small, apply t-test
3. Apply the appropriate formula for computing the value of the test statistic:
Z / t = Difference/Standard Error

Difference = Difference between two sample means

Standard Error = √ (σ12 / n1) + (σ22 / n2) (If population S.Ds are known)
2 2
Standard Error = √ (s1 / n1) + (s2 / n2) (If population S.D is unknown, but
samples are large)
Standard Error = √(n1s12+n2s22)/n1+n2 -2 x (1/n1 + 1/n2)
(If population S.Ds are unknown, and samples are small)
[Where σ 1= population S.D of sample 1, s1 = sample S.D of sample1, n1 = sample
size of sample 1; σ 2= population S.D of sample 2, s2 = sample S.D of sample2,
n2 = sample size of sample 2]
4. Specify the level of significance. If nothing is mentioned about the level of
significance, take 5%.
5. Fix the degree of freedom:
For Z-test, d.f = infinity; For t-test, d.f = n + n2 -1
 68

6. Locate the table value (critical value) of the test statistic at specified level of
significance and fixed degree of freedom.
7. Compare the calculated value of test statistic with the table value and decide
whether to accept or reject the null hypothesis. If calculated value of the test
statistic is numerically less than the table value, the null hypothesis is accepted. If
calculated value of the test statistic is numerically more than the table value, the
null hypothesis is rejected.

Qn: The mean yield of wheat from District I was 210Kg per acre from a sample of 100 plots.
In another District II, the mean yield was 200 Kg per acre from a sample of 150 plots.
Assuming that the S.D of yield of the entire State was 11 Kg, test whether there is any
significant difference between the mean yields of the crop in the two districts.

Sol:
District I District II
n1 = 100 n2 = 150
x̄1 = 210 x̄2 = 200
σ 1 = 11 σ 2 = 11

H0 : There is no significant difference between two sample means( i.e; μ1 = μ2)

H1 : is no significant difference between two sample means( i.e; μ1 ≠ μ2)

Since population S.Ds are given, the test statistic applicable here is Z-test.
Z = Difference / S E
Difference = x̄1 – x̄2 = 210 – 200 = 10
SE = √(σ12 / n1) + (σ22 / n2) (Population S.Ds are known. For the entire State SD is 11).
= √(112 / 100) + (112 / 150) = √1.21+0.81 = √2.02 = 1.42
Z = 10/1.42 = 7.04

Level of significance = 5%

Degree of freedom = infinity

Table value of Z at 5% level of significance and infinity degrees of freedom = 1.96

Since the calculated value of Z is more than the table value, H0 is rejected. We accept H1. So
we may conclude that there is significant difference in the mean yields of crops in two
districts.

Qn: Electric bulbs manufactured by X Ltd. and Y Ltd. gave the following results:

Particulars X Ltd Y Ltd

Number of bulbs used 100 100
Mean Life in Hours 1300 1248
Standard Deviation 82 93
State whether there is any significant difference in the life of bulbs of the two makes.
 69

Sol: H0 : There is no significant difference between two sample means( i.e; μ1 = μ2)
H1 : is no significant difference between two sample means( i.e; μ1 ≠ μ2)

Since population S.Ds are unknown but samples are large, the test statistic applicable
here is Z-test.
Z = Difference / S E
Difference = x̄1 – x̄2 = 1300 – 1248 = 52
SE = √(s12 / n1) + (s22 / n2) (Population S.Ds are unknown)
= √(822 / 100) + (932 / 100) = √67.24+86.49 = √153.73 = 12.4
Z = 52/12.4 = 4.19

Level of significance = 5%

Degree of freedom = infinity

Table value of Z at 5% level of significance and infinity degrees of freedom = 1.96

Since the calculated value of Z is more than the table value, H0 is rejected. We accept H1. (i.e;
μ1 ≠ μ2). So we may conclude that there is significant difference in the mean life of bulbs of
the two makes.

Qn: Two batches of same product are tested for their mean life. Assuming that lives of the
two products follow a normal distribution, test the hypothesis that the mean life is same for
both the batches, given the following information:

Batch Sample Size Mean life in hours S.D

A 10 750 12
B 8 820 14
Sol: H0 : There is no significant difference between two sample means( i.e; μ1 = μ2)
H1 : is no significant difference between two sample means( i.e; μ1 ≠ μ2)

Since population S.Ds are unknown and samples are small, the test statistic applicable
here is t-test.
t = Difference / S E
Difference = x̄1 – x̄2 = 820 – 750 = 70
SE = √(n1s12+n2s22)/n1+n2 -2 x (1/n1 + 1/n2) (Population S.Ds are unknown and
samples are small)
= √(10*122)+ (8*142) / 10+8-2) x (1/10 + 1/8)
= √3008/16 x 0.225 = √42.3 = 6.5
t = 70/6.5 = 10.77 (Calculated Value)

Level of significance = 5%

Degree of freedom = 10+8-2 = 16

Table value of t at 5% level of significance and 16 degrees of freedom = 2.12

 70

Since the calculated value of t is more than the table value, H0 is rejected. We accept H1. (i.e;
μ1 ≠ μ2). So we may conclude that the lives of products produced in two batches are not
same.

Qn: In a test given to 2 groups of students, the marks obtained were as follows:

Group I 18 20 36 50 49 36 34 49 41
Group II 29 26 28 35 30 44 46
Test whether the group means are equal.

Sol: Here we have to find the Means and S.Ds of the two samples.

Computation of Mean and S D of two Groups

Group I Group II
X X -- x̄ (X -- x̄)2 X X -- x̄ (X -- x̄)2
18 -19 361 29 -5 25
20 -17 289 26 -8 64
36 -1 1 28 -6 36
50 13 169 35 1 1
49 12 144 30 -4 16
36 -1 1 44 10 100
34 -3 9 46 12 144
49 12 144
41 4 16
εX= 333 1134 ε X = 238 386

Mean of Group I = 333/9 = 37 Mean of Group II = 238/7 = 34

S.D of Group I = √ ε (X -- x̄)2 / n S.D of Group I = √ ε (X -- x̄)2 / n

= √1134/9 = √126 = √386/7 = √55.14

H0 : There is no significant difference between two sample means( i.e; μ1 = μ2)

H1 : is no significant difference between two sample means( i.e; μ1 ≠ μ2)

Since population S.Ds are unknown and samples are small, the test statistic applicable
here is t-test.
t = Difference / S E
Difference = x̄1 – x̄2 = 37 – 34 = 3
SE = √(n1s12+n2s22)/n1+n2 -2 x (1/n1 + 1/n2) (Population S.Ds are unknown and
samples are small)
= √(9*126)+ (7*55.14) / 9+7-2) x (1/9 + 1/7)
= √1510.98/14 x 0.254 = √27.41 = 5.24
t = 3/5.24 = 0.573 (Calculated Value)

Level of significance = 5%
 71

Degree of freedom = 9+7-2 = 14

Table value of t at 5% level of significance and 14 degrees of freedom = 2.145

Since the calculated value of t is less than the table value, H0 is accepted. (i.e; μ1 = μ2). So
we may conclude that the difference in the group means are not significant. They are equal.

TESTING OF SIGNIFICANCE OF THE DIFFERENCE IN CASE OF DEPENDENT

SMPLES (PAIRED OBSERVATIONS)

Here the observations in one sample are some way related to the observations in the other.

Therefore they are called paired observations. The test statistic applicable here is t-test.

Procedure:

1. Set up H0 and H1
H0 : There is no significant difference between samples
H1 : There is significant difference between samples
2. Decide test statistic:
Since the paired data are comparatively less, the test statistic applicable here is always
t-test.
3. Apply the appropriate formula for computing the value of the test statistic.

t = d/SE

Where:

d Arithmetic mean of the difference between the values

SE S/√n-1 [s = standard deviation of the difference]

4. Specify the level of significance. Take 5%, if nothing is mentioned in the question.
5. Fix the degree of freedom. d.f = n – 1 , where n= Number of pairs of observations.
6. Locate the critical value of the test statistic (t-test) at specified level of significance
and fixed degree of freedom.
7. Compare the calculated value of test statistic with the table value and decide whether
to accept or reject the null hypothesis. If calculated value of the test statistic is
numerically less than the table value, the null hypothesis is accepted. If calculated
value of the test statistic is numerically more than the table value, the null hypothesis
is rejected.
Qn: The marks scored by 10 students, before and after providing special coaching, are
given in the following table:
Before 67 24 57 55 63 54 56 68 33 43
After 70 38 58 58 56 67 68 72 42 38
Test whether there is any significant difference in their performance.
Sol: H0 : There is no significant difference between samples
 72

H1 : There is significant difference between samples

Test statistic applicable here is t-test.

t = d/SE

Computation of men and standard deviation of the difference between the values
Score (Before) Score (After) Difference (d) d2
67 70 3 9
24 38 14 196
57 58 1 1
55 58 3 9
63 56 -7 49
54 67 13 169
56 68 12 144
68 72 4 16
33 42 9 81
43 38 -5 25
εd= 47 εd2 = 699

Arithmetic mean of d values = 47/10 = 4.7

S D of d values = √εd2/n -- (εd/n)2 = √699/10 -- (47/10)2 = √47.81 = 6.91

SE = 6.91/√10-1 = 6.91/3 = 2.3

t = 4.7/ 2.3 = 2.04

Level of significance = 5%

Degree of freedom = 10—1 = 9

Table value (critical value) of t at 5% level of significance and 9 degree of freedom is 2.262.

Since the calculated value of t is less than the critical value, the null hypothesis is accepted.
So, we may conclude that there is no significant difference in the performance of the
students.

TESTING OF GIVEN POPULATION PROPORTION

This type of testing of hypothesis is used to test whether there is any significant difference
between the sample proportion and the given population proportion.

Procedure:
 73

1. Set up H0 and H1 :
H0 : There is no significant difference between sample proportion and population
proportion ( i.e; H0 : P = P0)
H1 : There is significant difference between sample proportion and population
proportion ( i.e; H0 : P ≠ P0)
2. Decide the test statistic:
The test statistic applicable here is Z-test
3. Apply appropriate formulae for computing the value of Z ( i.e; calculated value):
Z = Difference / S E ie; Z = ( p -- P) / S E
Where p = sample proportion, P = Population proportion
S E = √ PQ / n
4. Decide the level of significance (Take 5%, if nothing is mentioned in the question).
5. Fix the degree of freedom ( Infinity d.f)
6. Locate the table value of Z at specified level of significance and fixed degree of
freedom.
7. Compare the calculated value of Z with the table value and decide whether to accept
or reject the null hypothesis. If calculated value of Z is numerically less than the table
value, the null hypothesis is accepted. If calculated value of Z is numerically more
than the table value, the null hypothesis is rejected.

Qn: It is found that out of 500 units of a product produced by a machine, 30 are
defectives. Test whether the machine produces 2% defective items on an average.
Sol:
H0 : There is no significant difference between sample proportion and population
proportion ( i.e; H0 : P = 0.02)
H1 : There is significant difference between sample proportion and population
proportion ( i.e; H0 : P ≠ 0.02)

Z = ( p -- P) / S E
P = 0.02, p = 30/500 = 0.06, Q = 1 – Q =1 - 0.02 = 0.98, n = 500
S E = √ PQ / n = √ 0.02 x 0.98 / 500 = √0.0196/500 = √0.0000392 = 0.0063
∴ Z = (0.06 – 0.02) / 0.0063 = 0.04/0.0063 = 6.349
Level of significance = 5%
Degree of freedom = infinity
Table value of Z at 5% level of significance and infinity degree of freedom is 1.96
Since the calculated value of Z is more than the table value, null hypothesis is rejected.
We accept alternative hypothesis. P ≠ 0.02. So, it is not possible to think that the
machine produces 2% defective items.
 74

TESTING OF THE SIGNIFICANCE OF THE DIFFERENCE BETWEEN TWO

SAMPLE PROPORTIONS
This testing of hypothesis is used to test whether the difference between two sample
proportions are significant or not. If the difference is not significant, they are treated as equal;
or we may think that the two samples are drawn from the same population.
Procedure:
1. Set up H0 and H1
H0 : There is no significant difference between two sample proportions( i.e; p1 = p2)
H1 : is no significant difference between two sample proportions( i.e; p1 ≠ p2)
2. Decide the test statistic:
The test statistic applicable here is Z-test.
3. Apply the appropriate formula for computing the value of the test statistic:
Z = Difference/Standard Error
i e; Z = p1 – p2 / SE
where p1 and p2 are the proportions of two samples
S E = √p0q0 [(1/n1) + (1/n2)] , where p0 = (n1p1 + n2p2)/(n1 + n2), q0= 1-- p0

4. Specify the level of significance. If nothing is mentioned about the level of

significance, take 5%.
5. Fix the degree of freedom (d f = infinity)
6. Locate the table value (critical value) of the test statistic at specified level of
significance and fixed degree of freedom.
7. Compare the calculated value of test statistic with the table value and decide
whether to accept or reject the null hypothesis. If calculated value of the test
statistic is numerically less than the table value, the null hypothesis is accepted. If
calculated value of the test statistic is numerically more than the table value, the
null hypothesis is rejected.
Qn: In a sample of 1000 people selected from District X, 450 were regular drinkers of coffee.
In another sample of 800 people drawn from District Y, 400 were regular drinkers of coffee.
Test whether there is significant difference between the two districts, regarding the coffee
drinking habit of people.
Sol:
H0 : There is no significant difference between two districts regarding the coffee drinking
habits of people ( i.e; p1 = p2)
H1 : is There is no significant difference between two districts regarding the coffee drinking
habits of people ( i.e; p1 ≠ p2)
The test statistic applicable here is Z-test.
Z = Difference/Standard Error
i e; Z = p1 – p2 / SE
p1 = 450/1000 = 0.45, p2 = 400/800 = 0.5
S E = √p0q0 [(1/n1) + (1/n2)] , n1 = 1000 , n2 = 800
p0 = (1000 x 0.45 + 800 x 0.5)/(1000 + 800) = 850/1800 = 0.472
 75

q0= 1— 0.472 = 0.528

∴ SE = √0.472 x 0.528 [(1/1000) + (1/800)] = √0.249 (0.001 + 0.00125)
= √0.249 x 0.00225 = √0.00056 = 0.0237
∴ Z = (0.5 – 0.45) / 0.0237 = 0.05/0.0237 = 2.11

Level of significance = 5%.

Fix the degree of freedom = infinity
Table value (critical value) of Z at 5% level of significance and infinity fixed degree
of freedom is 1.96
Since the calculated value of Z is more than the table value, null hypothesis is
rejected. Alternative hypothesis is accepted. p1 ≠ p2. So, we may conclude that
there is significant difference between the two districts regarding the coffee
drinking habits of people.

REVIEW QUESTIONS:

1. What do you mean by inferential analysis?

2. What do you understand by sampling distributions?
3. What are the two branches of inferential analysis?
4. What do you mean by hypothesis?
5. What is the difference between parameter and statistic?
6. What do you mean by Standard Error? What are its uses?
7. What are the differences between standard deviation and standard error?
8. What do you mean by parametric tests?
9. What do you mean by non-parametric tests?
10. What is type I error?
11. What is Type II error?
12. What do you mean by power of a test
13. What is meant by critical region and acceptance region?
14. What is one tailed test?
15. What is two-tailed test?
16. What do you mean by dependent sample?
17. Explain the general procedure for testing of hypothesis.
18. A random sample of 10 bottles of filled in by an automatic machine gave the following
weights in kilogram:
2.05, 2.01, 2.04, 1.96, 2.01, 1.98, 1.99, 1.98, 2.04, 2.02
Can we accept at 5% level of significance, the claim that the average weight of the tin is 2 Kg.
19. From the following data, test whether there is significant difference between two samples:
 76

Sample I 25 32 30 32 24 14 32
Sample II 24 34 30 22 42 31 40 35 32 30
20. In a sample of 600 people in Bihar 336 are coffee drinkers and the rest are tea drinkers. Can
we assume that both coffee and tea are equally popular in the State at 1% level of
significance?
21. In a sample of 900 men from a certain large city 675 were found to be smokers. In a random
sample of 1350 men from another large city 675 were found to be smokers. Do the data
indicate that the cities are significantly different in respect of the prevalence of smoking
among men?
22. A sample of size 50 has S.D of 10.5. Can you contradict the hypothesis that the population
S.D. is 12?
 77

CHAPTER 11

CHI-SQUARE TEST
What is Chi-Square Value?

The word “Chi-square” is denoted by the symbol, χ2. Chi-square is a value (quantity)
which describes the magnitude of the difference between observed frequencies and expected
frequencies.

Chi-Square Test

Chi-square test is a statistical test used to test the significance of the difference between
observed frequencies and the corresponding theoretical frequencies (expected frequencies) of
a distribution, without any assumption about the nature of distribution of the population. This
is the most popular widely used non-parametric test. It was developed by Prof. Karl Pearson.

Uses of Chi-Square Test (Applications of Chi-Square Test)

Chi-Square test is mainly used for the following purposes:

1. Used to test goodness of fit: As a test for goodness of fit, χ2 test can be used to test
how far the theoretical frequencies fit to the observed frequencies.
2. Used to test independence: As a test of independence, χ2 test is used to test whether
the attributes of a sample are associated or not.
3. Used to test homogeneity: As a test of homogeneity, χ2 test is used to test whether
different samples are homogeneous as far as a particular attribute is concerned.
4. Used to test population variance: Here, Chi-square test is used for testing the given
population variance when the sample is small. In other words, it used to test whether
there is any significant difference between sample variance and population variance.
Here, the test statistic value (Chi-square value) is obtained by using the following
formulae (ns2/σ2).

Conditions for applying Chi-square Test

1. The total frequencies(N) must be at least 50

2. Expected frequencies of less than 5 must be pooled with the preceding or succeeding
frequency so that the expected frequency is 5 or more.
3. The distributions should be of original units. They should not be of proportions or
percentages.

Testing of Goodness of Fit

Procedure:

1. Set up H0 and H1
H0 : There is goodness of fit between observed frequencies and expected frequencies.
H0 : There is no goodness of fit between observed frequencies and expected
 78

frequencies.
2. Decide the test statistic. Here, the test statistic is Chi-Square test.
3. Apply the appropriate formula:
χ2 = [(O—E)2/E]
where o = Observed frequencies and E = Expected frequencies
4. Specify the level of significance. If nothing is mentioned, take 5% level of
significance.
5. Fix the degree of freedom. Degree of freedom = n – r – 1
Where n = number of pairs of observations
r = number of parameters computed from the given data to find the expected
frequencies.
6. Obtain the table value of Chi-square at specified level of significance and fixed
degree of freedom.
7. Compare the actual value of Chi-Square with the table value and decide whether to
accept or reject the null hypothesis. If calculated value is less than the table value, null
hypothesis is accepted and otherwise it is rejected.

Qn: The numbers of road accidents per week in a certain city were as follows:

12, 8, 20, 2, 14, 10, 15, 6, 9, 4

Are these frequencies in agreement with the belief that the accidents occurred were
the same during the 10 week period?

Sol:

H0 : There is goodness of fit between observed frequencies and expected frequencies.

H0 : There is no goodness of fit between observed frequencies and expected
frequencies.

χ2 = [(O—E)2/E]
Here the Observed values (Actual values) are 12, 8, 20, 2, 14, 10, 15, 6, 9 and 4.

i.e; O = 12, 8, 20, 2, 14, 10, 15, 6, 9, 4

If accidents occurred are same, then the number of accidents per week which we may
expect is 10 (i.e; the average of the given values).

i.e; E = 10

Now we can find the value of Chi-square as follows:

Computation of Chi-square Value

Observed Expected (O – E)2 (O – E)2 / E
Values (O) Values (E)
12 10 4 0.4
8 10 4 0.4
20 10 100 10.0
 79

2 10 64 6.4
14 10 16 1.6
10 10 0 0.0
15 10 25 2.5
6 10 16 1.6
9 10 1 0.1
4 10 36 3.6
χ2 = 26.6

Calculated Value of χ2 = 26.6

Level of significance = 5%

Degree of Freedom = n -- r -- = 10 – 0 -- 1 = 9

Table value of χ2 at 5% level of significance and 9 d.f is 16.919.

Since calculated value is more than the table value, null hypothesis is rejected. We
accept alternative hypothesis. So we may conclude that the given figures do not agree
with the belief that accident occurred were same during the 10 weeks period.

Qn: The principal of a college made a sample analysis of an examination result of

200 students. It was found that 24 students had got first class, 62 second class, 68
third class and the rest were failed. Are these figures commensurate with the general
examination result which is in the ratio of 2:3:3:2 for various categories respectively.

Sol:

H0 : There is goodness of fit between the given figures and the figures expected in
general examination
H0 : There is no goodness of fit between the given figures and the figures expected in
general examination

χ2 = [(O—E)2/E]
Here the Observed values (Actual values) for first, second, third and failed categoris
of students are respectively 24, 62, 68 and 46.

If results are in the ratio of 2:3:3:2, then the number of students for above categories
may be expected as follows:

First Class 200 x 2/10 40

Second Class 200 x 3/10 60
Third Class 200 x 3/10 60
Failed 200 x 2/10 40
Total 200
So the E Values are 40, 60, 60 and 40.

Now we can find the value of Chi-square as follows:

 80

Computation of Chi-square Value

Observed Expected (O – E)2 (O – E)2 / E
Values (O) Values (E)
24 40 36 0.900
62 60 64 1.067
68 60 4 0.067
46 40 256 6.400
χ2 = 8.434

Calculated Value of χ2 = 8.434

Level of significance = 5%

Degree of Freedom = n -- r -- = 4 – 0 -- 1 = 3

Table value of χ2 at 5% level of significance and 9 d.f is 7.815.

Since calculated value is more than the table value, null hypothesis is rejected. We
accept alternative hypothesis. So we may conclude that the given figures do not
commensurate with the general examination result which is in the ratio of 2:3:3:2.

Testing of Independence

Procedure:

1. Set up H0 and H1
H0 : There is independence between observed frequencies and expected frequencies.
H0 : There is no independence between observed and expected frequencies.
2. Decide the test statistic. Here, the test statistic is Chi-Square test.
3. Apply the appropriate formula:
χ2 = [(O—E)2/E]
where o = Observed frequencies and E = Expected frequencies
Here E values are obtained by using the following formula:
E Value = [(Row Total x Column Total)/Grand Total]
E Values are computed by preparing a table called Contingency Table.
4. Specify the level of significance. If nothing is mentioned, take 5% level of
significance.
5. Fix the degree of freedom. Degree of freedom = (r – 1) x (c – 1)
Where r = number of rows; c = number of columns
6. Obtain the table value of Chi-square at specified level of significance and fixed
degree of freedom.
7. Compare the actual value of Chi-Square with the table value and decide whether to
accept or reject the null hypothesis. If calculated value is less than the table value, null
hypothesis is accepted and otherwise it is rejected.
 81

Qn: From the following data, can you say that there is relation between the habit of
smoking and literacy:

Smokers Non-smokes
Literates 83 57
Illiterates 45 68

Sol:

H0 : There is independence between smoking habit and literacy.

H0 : There is no independence between smoking habit and literacy.

χ2 = [(O—E)2/E]
Here the Observed values (Actual values) are 83, 57, 45 and 68.

The E Values corresponding to the above ‘O’ values can be found out by preparing a 2
X 2 contingency table:
2 X 2 Contingency Table
Smokers Non-smokes Total
Literates [(83+57) x (83+45)] /253 (140 x 125) / 253
= 71 = 69 140
Illiterates (113 x 128) /253 (113 x 125) / 253
= 57 = 56 113
Total 128 125 253
So, the E values are 71, 69, 57 and 56.

Computation of Chi-square Value

Observed Expected (O – E)2 (O – E)2 / E
Values (O) Values (E)
83 71 144 2.03
57 69 144 2.09
45 57 144 2.53
68 56 144 2.57
χ2 = 9.22
Calculated Value of χ2 = 9.22

Level of significance = 5%

Degree of Freedom = (2 – 1) x (2 – 1) = 1 x 1 = 1

Table value of χ2 at 5% level of significance and 1 d.f is 3.841.

Since calculated value is more than the table value, null hypothesis is rejected. We
accept alternative hypothesis. So we may conclude that there is no independence
between smoking habit and literacy. In other words, smoking habit and literacy are
related.
 82

Qn: In a sample study about the tea drinking habit in a town, following data are
observed in a sample of size 200.

46%nwerenmale, 26% were tea drinkers and 17% were male tea drinkers.

Is there any association between gender and tea habits?

Sol:

H0 : There is independence between gender and tea drinking habits.

H0 : There is no independence between gender and tea drinking habits.

χ2 = [(O—E)2/E]
Here all the Observed values (Actual values) are not directly given in the question.
So, we have to find the missing figures with the help of a 2 x 2 contingency table:

2 X 2 Contingency Table ( ‘O’ values)

Tea drinkers Non-tea Total
drinkers
Male (200 x 17) / 100 = 34 58 (200 x 46)/100 = 92
Female = 18 90
= 108
Total (200 x 26) / 100 = 52 148 200
“O” values are 34, 58, 18 and 90.

The E Values corresponding to the above ‘O’ values can be found out by preparing a 2
X 2 contingency table:
2 X 2 Contingency Table ( ‘E’ values)
Tea drinkers Non-tea drinkers Total
Male (92 x 52) / 200 = 24 (92 x 148) / 200 = 68 92
Female (108 x 52)/200 = 28 (108 x 148)/200 = 80 108
Total 52 148 200

So, the ‘E’ values are 24, 68, 28 and 80.

Computation of Chi-square Value

Observed Expected (O – E)2 (O – E)2 / E
Values (O) Values (E)
34 24 100 4.17
58 68 100 1.47
18 28 100 3.57
90 80 100 1.25
χ2 = 10.46
Calculated Value of χ2 = 10.46

Level of significance = 5%

Degree of Freedom = (2 – 1) x (2 – 1) = 1 x 1 = 1
 83

Table value of χ2 at 5% level of significance and 1 d.f is 3.841.

Since calculated value is more than the table value, null hypothesis is rejected. We
accept alternative hypothesis. So we may conclude that there is no independence
between gender and smoking habit. In other words, gender and smoking habit are
closely associated.

Testing of Homogeneity

Procedure:

1. Set up H0 and H1
H0 : There is homogeneity between the samples on the basis of the attribute.
H0 : There is no homogeneity between the samples on the basis of the attribute.
2. Decide the test statistic. Here, the test statistic is Chi-Square test.
3. Apply the appropriate formula:
χ2 = [(O—E)2/E]
where o = Observed frequencies and E = Expected frequencies
Here ‘E’ values are obtained by using the following formula:
‘E’ Value = [(Row Total x Column Total)/Grand Total]
‘E’ Values are computed by preparing a table called Contingency Table.
4. Specify the level of significance. If nothing is mentioned, take 5% level of
significance.
5. Fix the degree of freedom. Degree of freedom = (r – 1) x (c – 1)
Where r = number of rows; c = number of columns
6. Obtain the table value of Chi-square at specified level of significance and fixed
degree of freedom.
7. Compare the actual value of Chi-Square with the table value and decide whether to
accept or reject the null hypothesis. If calculated value is less than the table value, null
hypothesis is accepted and otherwise it is rejected.

Qn: In a diet survey the following results were obtained:

Hindus Muslim
No. of families drinking tea 124 16
No. families not drinking tea 56 10
Is there any difference between the communities in the matter of tea drinking?

Sol:

H0 : There is homogeneity between communities in the matter of tea drinking.

H0 : There is no homogeneity between communities in the matter of tea drinking.

χ2 = [(O—E)2/E]
Here the Observed values (Actual values) are 124, 16, 56, and 10
 84

The ‘E’ values corresponding to the above ‘O’ values can be found out by preparing a 2
X 2 contingency table:
2 X 2 Contingency Table
Smokers Non-smokes Total
No. of families (140 x 180) /206 = 122 (140 x 26)/206 = 18 140
drinking tea
No. of families (66 x 180) /206 = 58 (66 x 26) / 206 = 8
not drinking tea 66
Total 180 26 206
So, the ‘E’ values are 122, 18, 58 and 8.

Computation of Chi-square Value

Observed Expected (O – E)2 (O – E)2 / E
Values (O) Values (E)
124 122 4 0.033
16 18 4 0.222
56 58 4 0.069
10 8 4 0.500
χ2 = 0.824
Calculated Value of χ2 = 0.824

Level of significance = 5%

Degree of Freedom = (2 – 1) x (2 – 1) = 1 x 1 = 1

Table value of χ2 at 5% level of significance and 1 degree of freedom is 3.841.

Since calculated value is less than the table value, null hypothesis is accepted. So we
may conclude that there is homogeneity between communities in the matter of tea
drinking.

Testing of Variance

Procedure:

1. Set up H0 and H1
H0 : There is no significant difference between sample variance and population
variance.
H1 : There is significant difference between sample variance and population variance.
2. Decide the test statistic. Here, the test applicable is Chi-square test.
3. Apply the appropriate formula for computing the value of test statistic.
χ2 = ns2/σ2 , where n = sample size, s2 = sample variance, σ2 = population variance.
4. Specify the level of significance. Take 5%, unless specified otherwise.
5. Fix the degree of freedom. d.f = n –1.
6. Locate the table value of Chi-square at specified level of significance and fixed
degree of freedom.
 85

7. Compare the actual value of Chi-Square with the table value and decide whether to
accept or reject the null hypothesis. If calculated value is less than the table value, null
hypothesis is accepted and otherwise it is rejected.

Qn: A sample is drawn from a population which follows normal distribution. The size of
sample and S.D are respectively 10 and 5. Test whether this is consistent with the
hypothesis that the S D of the population is 5.3

Sol:

H0 : There is no significant difference between sample S.D and population S.D. (i.e; H0 :
S.D of population = 5.3)

H1 : There is significant difference between sample S.D and population S.D. (i.e; H1 : S.D
of population ≠ 5.3)

The test applicable is Chi-square test.

χ2 = ns2/σ2 , where n = 10, s2 = 52, σ2 = 5.32

= (10 x 52)/5.32 = 250/28.09 = 8.899

Specify the level of significance. Take 5%, unless specified otherwise.

Fix the degree of freedom. d.f = 10 –1 = 9

Table value of Chi-square at 5% level of significance and 9 degree of freedom is 16.919

Since calculated value is less than the table value, null hypothesis is accepted. So we may
conclude that there is no significant difference between sample S.D and population S.D.
The population S.D = 5.3

Qn: A sample group of 10 students are selected randomly from a class. Their weights (in
K.g) are 49, 40, 53, 38, 52, 47, 48, 45, 55, and 43. Can we say that the population
variance is 20 Kg?

Sol:

H0 : There is no significant difference between sample variance and population variance.

(i.e; H0 : Variance of population = 20)

H0 : There is significant difference between sample variance and population variance. (i.e;
H0 : Variance of population ≠ 20)

The test applicable is Chi-square test.

χ2 = ns2/σ2

Here, n = 10, σ2 = 20, Sample variance is to be computed from the given data.
 86

Computation of sample variance ( s2)

2
Weight ( X ) (X -- x̄) (X -- x̄)
49 2 4
40 -7 49
53 6 36
38 -9 81
52 5 25
47 0 0
48 1 1
45 -2 4
55 8 64
43 -4 16
2
X = 470 (X -- x̄) = 280
x̄ = 470/10 = 47.

2
Sample Variance (s2) = [ (X -- x̄) ]/n = 280/10 = 28.

χ2 = (10 x 28)/20 = 280/20 = 14.

Level of significance = 5%

Degree of freedom. (d.f) = 10 –1 = 9

Table value of Chi-square at 5% level of significance and 9 degree of freedom is 16.919

Since calculated value is less than the table value, null hypothesis is accepted. So we may
conclude that there is no significant difference between sample variance and population
variance. ∴ The population variance = 20 Kg.

REVIEW QUESTIONS:

1. What do you mean by Chi-square value?

2. What are the important uses of Chi-square test?
3. What do you mean by goodness of fit?
4. Explain the procedure for testing goodness of fit.
5. What do you mean by contingency table?
6. What are the important conditions for applying Chi-square test?
7. A die is thrown 150 times with the following results:
No. turned up 1 2 3 4 5 6
Frequency 19 23 28 17 32 31
Test the hypothesis that the die is unbiased.
 87

8. Following data are given:

Education
Gender
Middle High School College Total
Male 52 10 20 82
Female 44 12 26 82
Total 96 22 46 164
Can you say that education depends on gender?
 88

Chapter 12
ANALYSIS OF VARIANCE
Meaning of Analysis of Variance
The testing of hypotheses so far discussed consists of different sample groups which
do not exceed two. If there are three or more sample groups, the testing of equality of them
cannot be done in any of the methods which have already been discussed. The testing of
significance of the difference among three or more samples is generally done by using the
technique of analysis of variance. In case of analysis of variance, as part of testing procedure,
we have to prepare a separate statement called Analysis of Variance Table or ANOVA Table.
Therefore, this type of testing of hypothesis is also called analysis of variance. The test
statistic used for Analysis of Variance is F-test. F-test is a parametric test.

Types of Analysis of Variance

There are two types of Analysis of variance. They are:
1. One-way classification of data (One way analysis of variance)
2. Two-way classification of data (Two way analysis of variance)
One-way classification of data (One way analysis of variance)
In one way classification, observations are classified into different groups on the
basis of a single criterion. Suppose we want to study about the yield of a particular crop. You
know there are number of factors which influence the productivity of crops. If we undertake
this study to know the effect of quality of seed on the yield of crop, it is called one- way
analysis of variance. Here yield of crops based on different seed must be given in columns. In
other words, in case of one way analysis of variance, the samples must always be in columns.

Types of variances in One-way ANOVA

1. Variance between samples (Columns): This is the net result of the variation
different sample means from grand mean. Grand mean is the mean of all the
observations coming under all sample groups.
2. Variance within the sample: This is the net result of variations different items of
the sample from the respective sample means.
3. Variance about the sample: This is the sum of the variance between samples and
the variance within the sample.

Proforma of One-way ANOVA Table

One-way ANOVA Table
Source of Sum of Degree of Mean Sum of F-Ratio
variation Squares Freedom Squares
Between SSC = C–1 MSC = SSC/(C–1) [F= Larger variance ÷
Samples Smaller variance]
Within Sample SSE = N–C MSE = SSE/(N–C) F = MSC ÷ MSE, or
F = MSE ÷ MSC
Total SST = N –1
SSC = Sum of Squares between Columns (Samples)
 89

SSE = Sum of Square within Column (Sample)

SST = Sum of Square Total
MSC = Mean Sum of Squares between Columns (Samples)
MSE = Mean Sum of Squares within Column (Sample)
C = Number of Columns (Samples)

Procedure for carrying out One-way Analysis of variance

1. Set up H0 and H1.

H0 : There is no significant difference between samples
H1 : There is significant difference between samples
2. Decide the test statistic:
Test statistic applicable here is F-test
3. Apply the appropriate formula for computing the value of F-test.
F = Larger Variance ÷ Smaller Variance, i.e; MSC÷MSE or MSE÷MSC
(i) Find SST.
SST = Sum of square of all items – (T2/N)
Where T = Total of all observations, N = Total Number of observations
(T2/N) is generally called correction factor
(ii) Find SSC.
SSC = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)
Where εx1= sum of items in the first column
εx2= sum of items in the second column
n1 = number of items in the first column
n2 = number of items in the second column

(iii) Draw one-way ANOVA Table and enter the values of SST and SSC
(iv) Find the value of SSE. SSE = SST – SSC
(v) Find the degree of freedom in the third column as indicated in the proforma.
(vi) Find MSC. MSC = SSC ÷ (C–1)
(vii) Find MSE. MSE = SSE ÷ (N–C)
(viii) Find F–Ratio.
F= Larger variance ÷ Smaller variance; [i.e; F= MSC÷MSE, or F= MSE÷MSC]
4. Specify the level of significance. Take 5% if nothing is mentioned.
5. Fix the degrees of freedom. Here we have to fix a pair of d.f.
If ‘F’ is obtained by using F= MSC÷MSE, then pair of df is (d f of MSC, d f of MSE)
If ‘F’ is obtained by using F= MSE÷MSC, then pair of df is (d f of MSE, d f of MSC)
6. Obtain table value of F at specified level significance and fixed degree of freedom.
7. Compare the Calculated value of F with the Table value, and decide whether to accept
or reject the null hypothesis. If calculated value is less than the table value, H0 is
accepted. If calculated value is more than the table value, H0 is rejected.

Qn: Four varieties of a crop was grown on 3 plots, and the following yield was obtained. You
are required to test whether there is significant difference in the productivity of seeds:
 90

Variety of Seeds
Plot
P Q R S
I 10 7 8 5
II 9 7 5 4
III 8 6 4 4
Sol:

H0 : There is no significant difference in the productivity of seeds.

H1 : There is significant difference in the productivity of seeds.
Test Statistic applicable here is F-test
F = Larger Variance ÷ Smaller Variance, i.e; MSC÷MSE or MSE÷MSC
SST = 102 + 72 + 82 + 52 + 92 + 72 + 52 + 42 + 82 + 62 + 42 + 42 – (T2/N)
= 100+49+64+25+81+49+25+16+64+36+16+16 – (772/12)
= 541 – (5929/12) = 541 – 494 = 47
SSC = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)
= [(10+9+8)2÷3]+[(7+7+6)2÷3]+[(8+5+4)2÷3]+[(5+4+4)2÷3] – (T2/N)
= (729/3) + (729/3) + (729/3) + (729/3) – (772/12)
= (1587/3) – 494 = 529 – 494 = 35
One-way ANOVA Table
Source of Sum of Degree of Mean Sum of F-Ratio
variation Squares Freedom Squares
Between SSC = 35 C – 1= 4 –1 MSC = SSC/(C–1) [F= Larger variance ÷
Samples =3 35/3 = 11.67 Smaller variance]
Within Sample SSE = 12 N–C = 12 – 4 MSE = SSE/(N–C) F = MSC ÷ MSE, or
=8 12/8 = 1.5 = 11.67/1.5= 7.78
Total SST = 47 N –1= 12-
1=11
Level of significance = 5%
Degree of freedom = (3,8)
Table value of F at 5% level of significance and (3,8) degrees of freedom is 4.07
Since calculated value of F is more than the table value, H0 is rejected. We accept
alternative hypothesis. So we may conclude that there is significant difference in the
productivity of three varieties of seeds.

Qn: The following table shows the yield of 3 varieties. Perform analysis of variance and test
whether there is significant difference between varieties:
Plots
Varieties
A B C D E
I 30 27 42
II 51 47 37 48 42
III 44 35 41 36

Sol:
 91

Here, we are asked to test whether there is significant difference between varieties. But
varieties are given in rows, not in columns. In one way ANOVA, the samples must be in
columns. Therefore, we have to rearrange the given data so as to bring the samples in
columns as shown below:
Varieties
Plots
I II III
A 30 51 44
B 27 47 35
C 42 37 41
D 48 36
E 42 H0 : There is no significant difference in the
productivity of varieties.
H1 : There is significant difference in the productivity of varieties.
Test Statistic applicable here is F-test

F = Larger Variance ÷ Smaller Variance, i.e; MSC÷MSE or MSE÷MSC

SST = 30 + 51 + 44 + 27 + 47 + 35 + 422 + 372 + 412 + 482 + 362 + 422 – (T2/N)
2 2 2 2 2 2

= 900+2601+1936+729+2209+1225+1764+1369+1681+2304+1296+1764–
2
(480 /12)
= 19778 – 19200 = 578
SSC = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)
= [(30+27+42)2÷3]+[(51+47+37+48+42)2÷5]+[(44+35+41+36)2÷4] – (4802/12)
= (9801/3) + (50625/5) + (24336/4) – (19200)
= 3267 + 10125 + 6084 – 19200 = 19476 – 19200 = 276

One-way ANOVA Table

Source of Sum of Squares Degree of Mean Sum of F-Ratio
variation freedom Squares
Between SSC= 276 3-1=2 MSC=276/2=138 F=138/33.56
Samples = 4.112
Within SSE=302 12-3=9 MSE=302/9=33.56
Sample
Total SST=578 12-1=11
Level of significance = 5%
Degree of freedom = (2,9)
Table value of F at 5% level of significance and (2,9) degrees of freedom is 4.26
Since calculated value of F is more than the table value, H0 is accepted. So we may
conclude that there is no significant difference in the productivity of three varieties.
 92

Two-way classification of data (Two way analysis of variance)

In two way classification, observations are classified into different groups on the basis
of two criteria. Consider the example mentioned in one-way classification. If we study the
effect of both the quality of seeds and the type of fertilizers on the productivity of crop, the
data are to be classified on the basis of two criteria, namely type of seed and type of fertilizer.
This is called two-way analysis of variance. In case of two-way analysis of variance, we need
not make any kind of rearrangement in the given data. Since two criteria are considered, here,
there will be two sets of hypotheses.

Types of variances in Two-way ANOVA

1. Variance between samples (Columns): This is the net result of the variation
different sample means (in respect of columns) from grand mean. Grand mean is
the mean of all the observations coming under all sample groups.
2. Variance between rows: This is the net result of the variation different sample
means (in respect of rows) from grand mean. Grand mean is the mean of all the
observations coming under all sample groups.
3. Variance within the sample (Residual): This is the net result of variations
different items of the sample from the respective sample means.
4. Variance about the sample: This is the sum of the variance between columns,
variance between rows and the variance within the sample (residual)

Proforma of Two-way ANOVA Table

One-way ANOVA Table
Source of Sum of Degree of Mean Sum of Squares F-Ratio
variation Squares Freedom
Between SSC = c–1 MSC = SSC/(c–1) FC = MSC ÷ MSE, or
Columns = MSE ÷ MSC
Between Rows SSR = r–1 MSR = SSR/(r–1)
FR = MSR ÷ MSE, or
Within Sample SSE = (c – 1)x(r – 1) MSE=SSE/(c-1)x(r–1)
= MSE ÷ MSR

Total SST = N –1

SSC = Sum of Squares between Columns

SSR = Sum of Squares between Rows
SSE = Sum of Square within Samples
SST = Sum of Square Total
MSC = Mean Sum of Squares between Columns
MSR = Mean Sum of Squares between Rows
MSE = Mean Sum of Squares within Samples
c = Number of Columns
r = Number of Rows
 93

Procedure for carrying out Two-way Analysis of variance

1. Set up H0 and H1.

H0 : There is no significant difference between samples (in respect of columns)
H1 : There is significant difference between samples (in respect of columns)
H0 : There is no significant difference between samples (in respect of rows)
H1 : There is significant difference between samples (in respect of rows)
2 Decide the test statistic:
Test statistic applicable here is F-test
3 Apply the appropriate formula for computing the values of F ratios.
FC = Larger Variance ÷ Smaller Variance. i.e; MSC÷MSE or MSE÷MSC
FR = Larger Variance ÷ Smaller Variance. i.e; MSR÷MSE or MSE÷MSR

(i)Find SST.
SST = Sum of square of all items – (T2/N)
Where T = Total of all observations, N = Total Number of observations
(T2/N) is generally called correction factor
(ii)Find SSC.
SSC = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)
Where εx1= sum of items in the first column
εx2= sum of items in the second column
n1 = number of items in the first column
n2 = number of items in the second column

(iii)Find SSR.

SSR = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)

Where εx1= sum of items in the first row
εx2= sum of items in the second row
n1 = number of items in the first row
n2 = number of items in the second row

(iv)Draw one-way ANOVA Table and enter the values of SST, SSC and SSR
(v)Find the value of SSE. SSE = SST – (SSC+SSR)
(vi)Find the degree of freedom in the third column as indicated in the proforma.
(vii)Find MSC. MSC = SSC ÷ (c–1)
(viii)Find MSR. MSR = SSR ÷ (r–1)
(ix)Find MSE. MSE = SSE ÷ [(c–1) x (r–1)]
(x)Find F–Ratios (i. e; FC and FR)
 94

FC= Larger variance ÷ Smaller variance; [i.e; F= MSC÷MSE, or F= MSE÷MSC]

FR= Larger variance ÷ Smaller variance; [i.e; F= MSR÷MSE, or F= MSE÷MSR]
4. Specify the level of significance. Take 5% if nothing is mentioned.
5. Fix the degrees of freedom. Fix a pair of d.f. in respect of FC and FR.
6. Obtain table value of FC and FR at specified level significance and fixed degree of
freedom.
7. Compare the Calculated value of Fc with the Table value, and decide whether to
accept or reject the null hypothesis. If calculated value is less than the table value,
H0 is accepted. If calculated value is more than the table value, H0 is rejected.
8. Compare the Calculated value of FR with the Table value, and decide whether to
accept or reject the null hypothesis. If calculated value is less than the table value,
H0 is accepted. If calculated value is more than the table value, H0 is rejected.

Qn: Following table shows the yield of crops using 3 varieties of seeds:

Varieties of Seeds
Plots
P Q R
I 6 7 8
II 4 6 5
III 8 6 10
IV 6 9 9

Test whether there is significant difference in the productivity of varieties of seeds.

Also test the significance of the difference between plots.
Sol:

H0 : There is no significant difference in the productivity of varieties of seeds.

H1 : There is significant difference in the productivity of varieties of seeds.

H0 : There is no significant difference in the productivity of plots.

H1 : There is no significant difference in the productivity of plots.

Test statistic applicable here is F-test

F = Larger Variance ÷ Smaller Variance

FC = Larger Variance ÷ Smaller Variance. i.e; MSC÷MSE or MSE÷MSC

FR = Larger Variance ÷ Smaller Variance. i.e; MSR÷MSE or MSE÷MSR

SST = Sum of square of all items – (T2/N)
= 62 + 42 + 82 + 62 + 72 + 62 + 62 + 92 + 82 + 52 + 102 + 92 – (842/12)
=36+16+64+36+49+36+36+81+64+25+100+81– (7056/12)
= 624 – 588 = 36
 95

SSC = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)

= (242 /4) + (282/4) + (322/4) – 588
= (576/4) + (784/4) + (1024/4) – 588
= (2384/4) – 588 = 596 – 588 = 8
SSR = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)
= (212 /3) + (152/3) + (242/3) + (242/3) – 588
= (441/3) + (225/3) + (576/3) + (576/3) – 588
= (1818/3) – 588 = 606 – 588 = 18

One-way ANOVA Table

Source of Sum of Degree of Mean Sum of F-Ratio
Squares Freedom Squares
variation
Between SSC =8 3 – 1= 2 MSC = 8/2 = 4 FC = MSC ÷ MSE
Columns = 4/1.67 = 2.395
Between SSR =18 4–1=3 MSR = 18/3 = 6
FR = MSR ÷ MSE
Rows
= 6/1.67 = 3.593
Within SSE =10 (3 – 1)x(4 – 1) MSE=10/6 = 1.67
Sample =6
Total SST =36 12 –1 = 11

Between Columns:
Calculated value of FC = 2.396
Level of Significance = 5%
Degrees of freedom = (2,6)
Table value of FC at 5% level of significance and (2,6) degrees of freedom = 5.14
Since calculated value is less than the table value, null hypothesis is accepted. So we
may conclude that there is no significant difference in the productivity of three
varieties of seeds.
Between Rows:
Calculated value of FR = 3.593
Level of Significance = 5%
Degrees of freedom = (3,6)
Table value of FC at 5% level of significance and (3,6) degrees of freedom = 4.76
Since calculated value is less than the table value, null hypothesis is accepted. So we
may conclude that there is no significant difference in the productivity of plots.
 96

Coding Method
In analysis of variance, while preparing ANOVA table (both one-way and two-way),
at first, we have to find the values of SST, SSC, SSR, etc. But, if the individual observations
of the given data are of large values, the computation of SST, SSC, SSR, etc becomes a
tedious task. So, as to avoid this complication, we may apply coding method. Coding method
refers to the addition, subtraction, multiplication and division of individual observations of
the given data by a constant. The addition, subtraction, multiplication or division of all the
individual items by a constant will not affect the value of F.

Qn: The following table shows the number of units of a product produced by 5 workers using
4 different types of machines:
Machines
Workers
P Q R S
I 44 38 47 36
II 46 40 52 43
II 34 36 44 32
IV 43 38 46 33
V 38 42 49 39

You are required to test:

(1) Whether there is significant difference in the mean productivity of machines.
(2) Whether there is significant difference in the mean productivity of workers.
Sol:

Let us apply coding method by subtracting 45 from each observation of the given data. Then
we get;

Machines
Workers
P Q R S
I -1 -7 2 -9
II 1 -5 7 -2
III -11 -9 -1 -13
IV -2 -7 1 -12
V -7 -3 4 -6

H0 : There is no significant difference in the productivity of machines.

H1 : There is significant difference in the productivity of machines.

H0 : There is no significant difference in the productivity of workers.

H1 : There is no significant difference in the productivity of workers.

Test statistic applicable here is F-test

F = Larger Variance ÷ Smaller Variance

 97

FC = Larger Variance ÷ Smaller Variance. i.e; MSC÷MSE or MSE÷MSC

FR = Larger Variance ÷ Smaller Variance. i.e; MSR÷MSE or MSE÷MSR

SST = Sum of square of all items – (T2/N)
= -12 + 12 + -112 + -22 + -72 + -72 + -52 + -92 + -72 + -32 + 22 + 72+ -12 + 12 +
4 + -9 + -22 + -132 + -122 + -62– (-802/20)
2 2

=1+1+121+4+49+49+25+81+49+9+4+49+1+1+16+81+4+169+144+36–
(6400/20)
= 894 – 320 = 574

SSC = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)

= (-202 /5) + (-312/5) + (132/5) + (-422/5) – (-802/20)
= (400/5) + (961/5) + (169/5) + (1764/5) – 320
= (3294/5) – 588 = 658.8 – 320 = 338.8

SSR = [(εx1)2/n1] + [(εx1)2/n1] + [(εx1)2/n1] + ....................... – (T2/N)

= (-152 /4) + (12/4) + (-342/4) + (-202/4) + (-122/4) – (-802/20)
= (225/4) + (1/4) + (1156/4) + (400/4) + (144/4) – 320
= (1926/4) – 320 = 481.5 – 320 = 161.5

One-way ANOVA Table

Source of Sum of Degree of Mean Sum of F-Ratio
variation Squares Freedom Squares
Between SSC 4 – 1= 3 MSC = 338.8/3 = FC = MSC ÷ MSE
= 112.93/6.142
Columns =338.8 112.93
= 18.387
Between SSR 5–1=4 MSR = 161.5/4 =
Rows =161.5 FR = MSR ÷ MSE
40.375
= 40.375/6.142
Within SSE = (4 – 1)x(5 – 1) MSE=73.7/12 = = 6.574
73.7 = 12 6.142
Sample
Total SST =574 20 –1 = 19
Between Columns:
Calculated value of FC = 18.387
Level of Significance = 5%
Degrees of freedom = (3,12)
Table value of FC at 5% level of significance and (3,12) degrees of freedom = 3.49
 98

Since calculated value is more than the table value, null hypothesis is rejected.
Alternative hypothesis is accepted. So we may conclude that there is significant
difference in the mean productivity of machines.
Between Rows:
Calculated value of FR = 6.574
Level of Significance = 5%
Degrees of freedom = (4,12)
Table value of FC at 5% level of significance and (4,12) degrees of freedom = 3.26
Since calculated value is more than the table value, null hypothesis is rejected.
Alternative hypothesis is accepted. So we may conclude that there is significant
difference in the mean productivity of workers.

REVIEW QUESTIONS:

1. What do you mean by analysis of variance?

2. Explain the two types of analysis of variance.
3. What are the different types of variances in case of one way analysis of variance?
4. What are the different types of variances in case of two way analysis of variance?
5. Draw the proforma of one way ANOVA table.
6. Draw the proforma of two way ANOVA table.
7. Explain the hypothesis testing procedure in case of one way ANOVA.
8. Explain the hypothesis testing procedure in case of two way ANOVA.
9. What do you mean by coding method in analysis of variance?
10. Following table shows the scores attained by trainees under three different
instructional methods:

Methods Scores

I 84 71 84 76 85

II 85 76 88 86 90

III 81 68 73 71 82
Test whether there is significant difference in the scores under three methods.

11. A company had 4 salesmen P,Q,R and S, each of whom was sent for a period of one
moth to three types of areas, namely, urban area, rural area and semi-urban area. The
sales (in thousand rupees) achieved by the salesmen are shown in the following table:
 99

Salesmen
Area
P Q R S
Urban 80 80 60 100
Rural 30 30 70 30
Semi-urban 70 40 50 80
Carry out an analysis of variance and interpret the results.
 100

Chapter 13

NON-PARAMETRIC TESTS
Meaning:

A test which is not concerned with testing of parameters is called Non-parametric test.
Non-parametric test does not make any assumption about the nature of distribution.
Therefore, non-parametric tests are called distribution-free tests.

Situation where non-parametric tests are used

1. When hypothesis does not involve population parameter

2. When observations are not accurate as required for a parametric test.
3. When the researcher thinks that parametric test is not applicable.

Assumptions of Non-parametric tests

1. Samples are drawn randomly

2. Sample observations are independent
3. Observations are measured on ordinal or nominal scale
4. The variable is continuous
5. The probability density function of population is continuous

Advantages of Non-parametric tests

1. It is very simple and easy to apply the non-parametric tests

2. They can be applied when the observations are measured on ordinal or nominal scale
3. There is no assumption about the nature of population distribution
4. They can be used even if the sample is small
5. They have wide application in Psychometry, Sociology, Educational Statistics, etc.

Drawbacks of Non-parametric tests

1. They can be used only if the observations are measured on ordinal or nominal scale.
2. They cannot be used for estimating population parameters
3. The application of all non-parametric tests is not very simple.

Types of Non-parametric tests

1. Chi-square Test
2. Sign Tests
3. Signed Rank Test (Wilcoxon Matched Pairs Test)
4. Rank Sum Tests
5. One Sample Runs Test (Wald Wlfowitz’ Runs Test)
6. Kolmogrov - Smirnov Test (K-S Test)

Sign tests
 101

t-test is generally used when sample is small and there is an assumption that the population is
normal. Therefore, when sample is small but it is not possible to make an assumption about
the nature of population distribution, t-test cannot be applied. In such a case sign test is used.
In sign test, to find the value of test statistic, we use the proportion of signs (+ve or -ve
signs), not the numerical magnitude. That is why, the test is known as sign test. There are two
types of sign tests. They are (a) One sample sign test and (b) Two sample sign test.

One sample sign test

One sample sign test is used to test whether the sample belongs to a particular population.

Procedure:

1. Set up null and alternative hypotheses:

H0: There is no significant difference between sample mean and population mean (i.e; μ
= μ0)
H1: There is significant difference between sample mean and population mean ( i.e; μ =
μ0)
2. Decide the test statistic. The test statistic applicable is one sample sign test.
3. Use the appropriate formula for computing the value of test statistic
Test statistic = (p – P)/SE, where P = ½
p = proportion of + signs, (+ or – signs for each observation is determined by
subtracting 5 from each of them)
S E = √(PQ)/n where n = Total number of signs, Q = 1 – P
4. Specify the level of significance. Take 5%, if not mentioned.
5. Degree of freedom = infinity
6. Locate the table value of t-test.
7. Compare the calculated value with table value and decide whether to accept or reject
the hypothesis. If calculated value is less than the table value, null hull hypothesis is
rejected and otherwise, it is rejected.

Qn: Mr. A had to wait for following time (in minutes) for bus in 15 occasions:

9, 5, 6, 8, 3, 9, 8, 10, 7, 2, 6, 6, 7, 10 and 7 minutes. Use the sign test at 5% level of

significance to test the claim of bus that on the average Mr A has to wait 5 minutes.

Sol:

H0: There is no significant difference between sample mean and population mean (i.e;
μ = 5)
H1: There is significant difference between sample mean and population mean ( i.e; μ ≠
5)
The test statistic applicable is one sample sign test.
Test statistic = (p – P)/SE where p = proportion of + signs, P = ½
S E = √(PQ)/n
 102

Computation of proportion of + signs (p):

Waiting time (x) (X – 5)

9 +
5 .
6 +
8 +
3 –
9 +
8 +
10 +
7 +
2 –
6 +
6 +
7 +
10 +
7 +
No. of + signs = 12
Total number of signs (n) = 14
Total number of + signs = 12
Proportion of + signs (p) = (12/14) = 0.857
P = 0.5, Q = 1- 0.5 = 0.5
S E = √(0.5 x 0.5)/14 = √0.25/14 = √0.01786 = 0.1336
Test statistic = (0.857 – 0.5)/0.1336 = 2.672
Level of significance = 5%
Degree of freedom = infinity
Table value of ‘t’ at 5% level of significance and infinity degrees of freedom is 1.96
Since calculated value is more than the table value, null hypothesis is rejected. We
accept alternative hypothesis. So we may conclude that there is significant difference
between sample mean and population mean ( i.e; μ ≠ 5).

Two sample sign test (Paired sample sign test)

Two sample sign test is used to test whether two populations are identical. In case of two
sample sign test, each pair is replaced by +ve or --ve sign. If first vale in a pair is larger,
assign + ve sign to that pair, and otherwise assign – ve sign. Procedure the procedure is same
as in the case of one sign test.
Qn: The following are the scores obtained by 2 students in different tests:

Student I 7 10 14 12 6 9 11 13 7 6 10
Student II 10 13 14 11 10 7 15 11 10 9 8
Use the sign test at 1% level of significance to test the null hypothesis that on an average the
two students are identical.
Sol:

H0: There is no significant difference between students ( i.e; performance of the students are
identical)
H1: There is significant difference between students
 103

The test statistic applicable is two sample sign test.

Test statistic = (p – P)/SE where p = proportion of + signs, P = ½

S E = √(PQ)/n

Computation of proportion of + signs (p):

Student I (x) Student II (y) Sign for difference
7 10 +
10 13 +
14 14 .
12 11 –
6 10 +
9 7 –
11 15 +
13 11 –
7 10 +
6 9 +
10 8 –
No. of + signs = 6
Total number of signs (n) = 10
Total number of + signs =6
Proportion of + signs (p) = (6/10) = 0.6
P = 0.5, Q = 1- 0.5 = 0.5
S E = √(0.5 x 0.5)/10 = √0.25/10 = √0.025 = 0.1581
Test statistic = (0.6 – 0.5)/0.1581 = 0.1/0.1581 = 0.633
Level of significance = 1%
Degree of freedom = infinity
Table value of ‘t’ at 1% level of significance and infinity degrees of freedom is 2.576
Since calculated value is less than the table value, null hypothesis is accepted. So we may
conclude that the students are identical in their performance.

Signed Rank Test (Wilcoxon Matched Pairs Test)

Signed Rank Test is another important non-parametric test used to test whether matched
paired samples are identical or not. Here we use the signed ranks for testing. Wilcoxon
Matched Pairs Test is used differently depending upon following two situations:

(a) When the number of matched pairs ≤ 25, and

(b) When the number of matched pairs > 25.

Signed Rank Test (When the number of matched pairs ≤ 25)

Here, we find the difference of matched pairs and assign them ranks. Then ranks are
classified into two categories based on their respective signs. Then take the sun of two
categories of ranks. The minimum of the two is considered as the value of test statistic.

Procedure:
 104

1. Set up null and alternative hypotheses:

H0: There is no significant difference between samples
H1: There is significant difference between samples
2. Decide the test statistic. The test statistic applicable here is Wilcoxon matched pairs
test ( i.e; Wilcoxon’s T test)
3. Use the appropriate formula for computing the value of test statistic (Wilcoxon’s T
test)
T = Sum of Positive Ranks or Sum of Negative Ranks, whichever is less.
4. Specify the level of significance. Take 5%, if not mentioned.
5. Degree of freedom = n-1
6. Locate the table value of Wilcoxon’s T test.
7. Compare the calculated value with table value and decide whether to accept or reject
the hypothesis. If calculated value is less than the table value, null hull hypothesis is
rejected and otherwise, it is rejected.

Qn: The following table shows the details of number of units of a product produced by
two workers. Test whether there is significant difference between the performances of the
workers using Wilcoxon matched pairs test.

Worker P 73 43 47 53 58 47 52 58 38 61 56 56 43 55 65 75
Worker Q 51 41 43 41 47 32 24 58 43 53 52 57 44 57 40 68
Sol:

H0: There is no significant difference between samples

H1: There is significant difference between samples
Decide the test statistic. The test statistic applicable here is Wilcoxon matched pairs test
( i.e; Wilcoxon’s T test)
T = Sum of Positive Ranks or Sum of Negative Ranks, whichever is less.
Worker P Worker Q Difference (d) Rank of ldl Ranks of +veRanks of –
(P – Q) values ve
values
73 51 22 22 13 13
43 41 2 2 2.5 2.5
47 43 4 4 4.5 4.5
53 41 12 12 11 11
58 47 11 11 10 10
47 32 15 15 12 12
52 24 28 28 15 15
58 58 0 0 - -
38 43 -5 5 6 - -6
61 53 8 8 8 8
 105

56 52 4 4 4.5 4.5
56 57 -1 1 1 -1
34 44 -10 10 9 -9
55 57 -2 2 2.5 -2.5
65 40 25 25 14 14
75 68 7 7 7 7
Total of Signed Ranks 101.5 18.5
The calculated value of T = 101.5 or 18.5 whichever is lower.
∴ T value = 18.5
Level of significance = 5%
Degree of freedom = n-1, (n= number of vlues who have either + or –ve sign)
N = 15
Table value of Wilcoxon’s T test at 5% level of significance and 15 df = 25
Since calculated value is less than the table value, null hypothesis is accepted. So
we may conclude that there is no significant difference in he performances of
workers are P and Q.

Signed Rank Test (When the number of matched pairs > 25)

Here, we find the difference of matched pairs and assign them ranks. Then ranks are
classified into two categories based on their respective signs. Then take the total of two
categories of ranks. The test statistic is Z test.

Procedure:

1. Set up null and alternative hypotheses:

H0: There is no significant difference between samples
H1: There is significant difference between samples
2. Decide the test statistic. The test statistic applicable here is Wilcoxon matched pairs
test ( i.e; Z test)
3. Use the appropriate formula for computing the value of test statistic (Z test)
Z = [(T – μ)/σ ] where T = Sum of Positive Ranks or Sum of Negative Ranks,
whichever is less; μ = [n(n+1)]/4; σ = √ [n(n+1)(2n+1)]/24
4. Specify the level of significance. Take 5%, if not mentioned.
5. Degree of freedom = infinity
6. Locate the table value of Z.
7. Compare the calculated value with table value and decide whether to accept or reject
the hypothesis. If calculated value is less than the table value, null hull hypothesis is
rejected and otherwise, it is rejected.

Qn: The following are the marks obtained by 26 students before and after giving a special
coaching to them:
 106

Marks (before) : 70, 35, 21, 16, 75, 63, 70, 54, 77, 82, 68, 19, 13, 72, 78, 17, 24, 3, 45, 80, 15,
20, 58, 65, 35, 52.
Marks (after) : 79, 62, 90, 37, 35, 14, 26, 32, 90, 54, 85, 44, 83, 90, 92, 32, 34, 28, 34, 79, 35,
32, 62, 63, 30, 68.
Use the signed rank test to test at whether there is significant difference in the marks of
students before and after providing special coaching (∝ = 5%).

Sol:

H0: There is no significant difference in the marks of students before and after giving
special coaching.
H1: There is no significant difference in the marks of students before and after giving
special coaching.
The test statistic is Wilcoxon matched pairs test ( i.e; Z test)
Z = [(T – μ)/σ ] where T = Sum of Positive Ranks or Sum of Negative Ranks,
whichever is less; μ = [n(n+1)]/4; σ = √ [n(n+1)(2n+1)]/24

Marks Marks (after)Difference (d) Rank of ldl Ranks of +veRanks of –

(before) values ve
values
70 79 -9 9 6 6
35 62 -27 27 20 20
21 90 -69 69 25 25
16 37 -21 21 17 17
75 35 40 40 22 22
63 14 49 49 24 24
70 26 44 44 23 23
54 32 22 22 18 18
77 90 -13 13 10 10
82 54 28 28 21 21
68 85 -17 17 14 14
19 44 -25 25 19 19
13 83 -70 70 26 26
72 90 -18 18 15 15
78 92 -14 14 11 11
17 32 -15 15 12 12
24 34 -10 10 7 7
35 28 7 7 5 5
 107

45 34 11 11 8 8
80 79 1 1 1 1
15 35 -20 20 16 16
20 32 -12 12 9 9
58 62 -4 4 3 3
65 63 2 2 2 2
35 30 5 5 4 4
52 68 -16 16 13 13
Total of Signed Ranks 128 223
T = 128 (128 or 223, whichever is low)
μ = [n(n + 1)]/4, = 26(26+1)/4 = (26 x 27)/4 = 702/4 = 175.5
σ = √[n(n+1)(2n+1)]/24 = √[26(26+1)(52+1)]/24 = √(26 x 27 x 53)/24
= √37206/24 = √1550.25 = 39.373
∴ Z = (128 – 175.5)/39.373 = – 47.5/39.373 = – 1.21 = 1.21 (numerically)
Level of significance = 5%
Degree of freedom = infinity
Table value of Z at 5% level of significance and infinity degree of freedom = 1.96
Since calculated value is less than the table value, null hypothesis is accepted. So
we may conclude that there is no significant difference in the performances of
students before and after giving special coaching.

Rank Sum Tests

Rank sum tests are another type of tests used for testing whether the
populations are identical. Here, various samples are taken together and then ranks are
assigned. There are two important types of rank sum tests. They are (a) Wilcoxon-
Mann-Whitney Test (U- test), and (b) Kruskal – Wallis Test ( H- test).

Wilcoxon-Mann-Whitney Test (U- test):

This method is used when there are two group of samples. The testing procedure is:

1. Set up null and alternative hypotheses:

H0: There is no significant difference between two samples
H1: There is significant difference between two samples
2. Decide the test statistic. The test statistic applicable here is Wilcoxon Mann Whitney
test ( i.e; U- test)
3. Use the appropriate formula for computing the value of test statistic (Z test)
Test Statistic = [(μ–U)/SE ]
where μ = (n1.n2)/2
 108

U = U1 or U2 whichever is less.
U1 = n1.n2 + [n1(n1+1)]/2 – R1
U2 = n1.n2 + [n2(n2+1)]/2 – R2
R1 = Rank sum of Sample I
R2 = Rank sum of Sample II
n1 = Number of observations in Sample I
n2 = Number of observations in Sample II
SE= √ [n1.n2 (n1+ n2+1)]/12
4. Specify the level of significance. Take 5% unless specified otherwise.
5. Fix the degrees of freedom. df = infinity
6. Locate the table value of test statistic (i.e; Z test) at specified level of significance and
fixed degrees of freedom.
7. Compare the calculated value with table value and decide whether to accept or reject
the hypothesis. If calculated value is less than the table value, null hull hypothesis is
rejected and otherwise, it is rejected.

Qn: Apply Wilcoxon- Mann-Whitney Test to test whether the following samples come from
populations with same mean (i.e; they are identical):

Sample I 54 39 70 58 47 40 74 49 74 75 61 79
Sample II 45 41 62 53 33 45 71 42 68 73 54 73
Sol:

H0: There is no significant difference between two samples (i.e; they are identical)
H1: There is significant difference between two samples (i.e; they are not identical)
The test statistic is Wilcoxon Mann Whitney test ( i.e; U- test)
Value of Test Statistic = [(μ–U)/SE ]
where μ = (n1.n2)/2
U = U1 or U2 whichever is less.
U1 = n1.n2 + [n1(n1+1)]/2 – R1
U2 = n1.n2 + [n2(n2+1)]/2 – R2
R1 = Rank sum of Sample I
R2 = Rank sum of Sample II
SE= √ [n1.n2 (n1+ n2+1)]/12
Computation of Rank Sums
Values of Samples Rank Rank
together (ascending
order) sample I sample II
33 1 1
39 2 2
 109

40 3 3
41 4 4
42 5 5
45 6.5 6.5
45 6.5 6.5
47 8 8
49 9 9
53 10 10
54 11.5 11.5
54 11.5 11.5
58 13 13
61 14 14
62 15 15
68 16 16
70 17 17
71 18 18
73 19.5 19.5
73 19.5 19.5
74 21.5 21.5
74 21.5 21.5
75 23 23
79 24 24
Rank Sum R1=167.5 R2=132.5
U1 = 12 x 12 + [12(12+1)]/2 – 167.5 = 144 + 78 – 167.5 = 54.5
U2 = 12 x 12 + [12(12+1)]/2 – 132.5 = 144 + 78 – 132.5 = 89.5
U = 54.5 or 89.5 whichever is lower, ∴ U = 54.5
μ = (12 x 12)/2 = 144/2 = 72
SE = √[(12 x 12)(12 +12 + 1)]/12 = √(144 x 25)/12 = √300 = 17.32
∴ Test Statistic = (72 – 54.5)/17.32 = 17.5/17.32 = 1.011

Level of significance = 5%.

Degrees of freedom = infinity
 110

Table value of test statistic (i.e; Z test) at 5% level of significance and infinity degrees
of freedom = 1.96
Since calculated value is less than the table value, null hypothesis is accepted. So
we may conclude that there is no significant difference between two samples. Both
the samples come from populations with the same mean.

Kruskal – Wallis Test ( H- test):

Here, we tests whether three or more independent sample groups come from the
population having the same mean. The testing procedure is:
1. Set up null and alternative hypotheses:
H0: There is no significant difference between samples
H1: There is significant difference between samples
2. Decide the test statistic. The test statistic applicable here is Kruskal – Wallis test ( i.e;
H- test)
3. Use the appropriate formula for computing the value of test statistic.
Test Statistic H = [12/n(n+1)] x [εR12/n1+ εR22/n2)+........... ] – 3(n+1)
R1 = Rank sum of Sample I
R2 = Rank sum of Sample II
n1 = Number of observations in Sample I
n2 = Number of observations in Sample II
4. Specify the level of significance. Take 5% unless specified otherwise.
5. Fix the degrees of freedom. df = c–1
6. Locate the table value of test statistic (i.e; Chi-square test) at specified level of
significance and fixed degrees of freedom.
7. Compare the calculated value with table value and decide whether to accept or reject
the hypothesis. If calculated value is less than the table value, null hull hypothesis is
rejected and otherwise, it is rejected.

Qn: the sales figures of 4 salesmen are given below:

Salesmen Sales (Rupees in thousands)

P 171 182 157 148 162
Q 152 175 202 168 176
R 160 155 139 146 166
S 179 142 197 170 158
Test whether 4 salesmen have performed equally. Use Kruskal-Wallis Test at 1%.

Sol:
H0: There is no significant difference between salesmen
H1: There is significant difference between salesmen
The test statistic applicable here is Kruskal – Wallis test ( i.e; H- test)
Use the appropriate formula for computing the value of test statistic.
Test Statistic H = [12/n(n+1)] x [εR12/n1+ εR22/n2)+........... ] – 3(n+1)
 111

R1 = Rank sum of Sample I

R2 = Rank sum of Sample II
n1 = Number of observations in Sample I
n2 = Number of observations in Sample II
Computation of Rank Sums
Sales figures of 4 Ranks of Salesmen
salesmen Rank
(ascending order)
139 1 1
142 2 2
146 3 3
148 4 4
152 5 5
155 6 6
157 7 7
158 8 8
160 9 9
162 10 10
166 11 11
168 12 12
170 13 13
171 14 14
175 15 15
176 16 16
179 17 17
182 18 18
197 19 19
202 20 20
Rank Sums R1=53R2=68R3=30R4=59

H = [12/(n(n+1))] x [(εR12/n1)+ (εR22/n2)+........... ] – 3(n+1)

= 12/(20 x 21) x [(532/5)+(682/5)+(302/5)+(592/5) – 3(20+1)
= (12/420) x (561.8+924.8+180+696.2) – 63
= (0.02857 x 2362.8) – 63 = 67.5052 – 63 = 4.5052
Level of significance = 1%
 112

Degree of freedom = 4–1= 3

Table value of Chi-square at 5% level of significance and 3 d.f = 11.341
Since calculate vale is less than table value, null hypothesis is accepted. So we may
conclude that there is no significant difference between 4 salesmen. Their
performance are equal.

One Sample Runs Test (Wald Wolfiwitz’ Runs Test)

Runs test is used to test the randomness of a sample on the basis of the order in which
the observations are taken. A ‘run’ is a succession of identical items. This test was
designed by Wald Wolfowitz. The testing procedure is:
1. Set up null and alternative hypotheses:
H0: There is randomness
H1: There is no randomness
2. Decide the test statistic. The test statistic applicable here is Z-test.
3. Use the appropriate formula for computing the value of test statistic.
Z = (r – μ)/σ
where r = Number of runs; μ = [2n1n2/(n1+n2)] + 1

σ = √[2n1n2(2n1n2–n1–n2)]/ (n1+n2)2(n1+n2–1)

n1 = Number of first item in all the runs together

n2 = Number of second item in all the runs together

4. Specify the level of significance. Take 5% unless specified otherwise.

5. Fix the degrees of freedom. df = infinity
6. Locate the table value of Z at specified level of significance and infinity degrees
of freedom.
7. Compare the calculated value with table value and decide whether to accept or
reject the hypothesis. If calculated value is less than the table value, null hull
hypothesis is rejected and otherwise, it is rejected.

Qn: Test the randomness of following arrangement of students (Boys and Girls) in a class:

B,G,B,G,B,B,B,G,B,G,B,B,B,G,G,B,B,B,B,G,G,B,G,B,B,B,G,B,B,B,G,G,G,B,G,B,B,B
,G,B,G,B,B,B,B,G,G,B

Sol:
H0: There is randomness
H1: There is no randomness
The test statistic applicable here is Z-test.
Use the appropriate formula for computing the value of test statistic.
 113

Z = (r – μ)/σ
r = Number of runs
μ = [2n1n2/(n1+n2)] + 1

σ = √[2n1n2(2n1n2–n1–n2)]/ (n1+n2)2(n1+n2–1)

n1 = Number of first item in all the runs together

n2 = Number of second item in all the runs together
B,/G,/B,/G,/B,B,B,/G,/B,/G,/B,B,B,/G,G,/B,B,B,B,/G,G,/B,/G,/B,B,B,/G,/B,B,B,/G,G,
G,/B,/G,/B,B,B,/G,/B,/G,/B,B,B,B,/G,G,/B/

Number of runs (r) = 27

n1= 30
n2= 18
μ = [2n1n2/(n1+n2)] + 1 = [2 x 30 x 18/(30+18)] + 1
= (1080/48)+1 = 22.5 + 1 = 23.5

σ = √[2n1n2(2n1n2–n1–n2)]/ (n1+n2)2(n1+n2–1)
= √[2 x 30 x 18(2*30*18 – 30–18)]/ [(30+18)2(30+18–1)]

= √1080(1080-30-18)/(482 * 47) = √(1080 x 1032)/108288

= √1114560/108288 = √10.2926 = 3.208

∴ Z = (27 – 23.5)/3.208 = 3.5/3.208 = 1.091

Level of significance = 5%
Degree of freedom = infinity
Table value of Z at 5% level of significance and infinity d f = 1.96
Since calculate vale is less than table value, null hypothesis is accepted. So we may
conclude that the arrangement is made at random.

REVIEW QUESTIONS:

1. What do you mean by non-parametric tests?

2. What are the situations under which non-parametric tests are applied?
3. What are the important assumptions of non-parametric tests?
4. What are the important merits of non-parametric tests?
5. What are the important drawbacks of non-parametric tests?
6. What are the different types of non-parametric tests?
7. Distinguish between parametric tests and non-parametric tests.
8. What do you mean by one sample sign test?
9. Explain the hypothesis testing procedure under one sample sign test.
 114

10. Explain the hypothesis testing procedure under two sample sign test.
11. What do you mean by Wilcoxon matched pairs test?
12. Explain the hypothesis testing procedure of Wilcoxon matched pairs test.
13. What is meant by Wilcoxon Mann Whitney U-test?
14. Explain the hypothesis testing procedure of Wilcoxon Mann Whitney U-test.
15. What is meant by Kruskal-Wallis H-test?
16. Explain the hypothesis testing procedure of H-test.
17. What do you mean by one sample runs test?
18. Explain the hypothesis testing procedure under one sample runs test.
19. The following are the measurements of the breaking strength of a certain commodity:
173, 187,163, 172, 166, 163, 165, 160, 189, 161, 171, 158, 151, 169, 162, 163, 139,
172, 165 and 148. Use sign test to test the null hypothesis that mean breaking strength
of the commodity is 160.
20. A driver buys petrol either at station X or at station Y. the following arrangement
shows the order of the stations from which the driver bought petrol over a certain
period of time:
X, X, X, Y, X,Y, X,Y, X, Y, Y, Y, X, Y, Y, Y, X, Y, Y, X, Y, X, Y, X, Y, Y, X, Y,
Y, X, X,Y, X, Y, Y, Y, X, Y, X, X, X, Y, X, X, Y, X, X, X, X, Y.
 115

Chapter 14
SAMPLE SIZE DETERMINATION
Determination of size of sample is very important. If the sample size is very large, it
will be very difficult to manage the data. But, if the size is too small, the sample will not
represent the population, and the conclusion drawn may not be correct. Therefore, the size of
sample must be optimum.

Following are some of the important formulae commonly used for determining
sample size:

A. Sample Size Determination While Estimating Population Mean When

Population is Infinite

Sample Size (n) = (Zσ/e)2

where n = sample size, Z = table value, σ = S D of population
e = allowable difference between population mean and sample mean.
Qn: From the details given below, determine the sample size for estimating population
mean:
(a) Population S D = 15
(b) Confidence level = 99%
(c) Estimate should be within 6 units of the population mean.

Sol:
n = (Zσ/e)2
Population S D (σ) = 15
Allowable difference (e) = 6
Value of Z at 1% level of significance and infinity d f = 2.576
∴ n = [(2.576 x 15)/6]2 = (38.64/6)2 = 6.442 = 41.474
= 41

B. Sample Size Determination While Estimating Population Mean When

Population is Finite

Sample Size (n) = [Z2Nσ2] / {[(N-1)e2] + [Z2σ2]}

where Z = table value of Z; N = Size of population; σ = S D of Population
e = allowable difference between population mean and sample mean.
Qn: From the details given below, determine the sample size for estimating population mean:
(a) Population size = 5000
(b) Variance of the population = 4
(c) Estimate should be within 0.4 units of the true value of the population mean
 116

(d) Desired level of confidence = 99%

Sol:
n = [Z2Nσ2] / {[(N-1)e2] + [Z2σ2]}
Population Sixe (N) = 5000
Population S D = √4 = 2
Allowable difference (e) = 0.4
Value o at 1% level of confidence and infinity df = 2.576
∴ n = [2.5762 * 5000*22 ] / {[(5000-1)0.42] + [2.5762*22]}
= (6.6358*5000*4) / (799.84 + 26.543) = 132716/826.383
= 160.599 = 161
C. Sample Size Determination While Estimating Population Proportion When
Population is Infinite

Sample Size (n) = [Z2pq/ e2]

where Z = Table value of Z
p = sample proportion; q = (1 – p)
e = allowable difference between population proportion and sample
proportion.
Qn: It is decided to draw a sample from a population to estimate the percentage of defectives
within 2% of the true value with 95.5% confidence, on the basis of 3% defective in the
sample. What should be the sample size?

Sol:

n = [Z2pq/ e2]

p = 3% = 0.03; q = 1 – 0.03 = 0.97

e = 2% = 0.02

Value of Z at 4.5% level of significance and infinity df = 2.005

∴ n = [(2.0052 *0.03*0.97)/0.022] = 0.116983/0.0004 = 292.457

n = 292

D. Sample Size Determination While Estimating Population Proportion When

Population is Finite

Sample Size (n) = [Z2Npq] / {[(N-1)e2] + [Z2pq]}

Where Z = Table value of Z; p = sample proportion; q = (1 – p)
 117

N = Size of population
e = allowable difference between population proportion and sample
proportion.

Qn: It is decided to draw an optimal sample from a population of 5000 units to estimate the
percentage of defectives on the basis of 3% defectives in the sample within 0.05 units of its
true value. Level of confidence desired is 95%.

Sol:
Sample Size (n) = [Z2Npq] / {[(N-1)e2] + [Z2pq]}

N = 5000

p = 3% = 0.03

q = (1 – 0.03) = 0.97

e = 0.05

Table Value of Z at 5% level of confidence and infinity df = 1.96

∴ n = [1.962*5000*0.03*0.97] /{[(5000-1)0.052] + [1.962*0.03*0.97]}

= 558.9528/(12.4975 + 0.111791) = 558.9528/12.6093

= 44.33 = 44.

REVIEW QUESTIONS:

1. What do you mean by sample size?

2. What are the important formulae used for determining sample size while estimating
population mean?

3. What are the important formulae used for determining sample size while estimating
population proportion?
 118

Chapter 15

STATISTICAL ESTIMATION
Statistical estimation is one of the important branches of Statistical inferences. It is
concerned with estimation of population parameters with the help of samples drawn from that
population. The accurate value of population parameter can be computed only by an
exhaustive study of the population. But, it is infeasible to collect date from each and every
element of the population. Therefore, we estimate that population parameters through sample.
This is the actual process of statistical estimation.

Two types of estimates are generally used for estimating population parameter. They
are (a) Point Estimate and (b) Interval Estimate.

Point Estimation
If a single statistic is used as an estimate of an unknown parameter, it is called point estimate
of that parameter. Eg; when the particular value of the sample mean is called the “estimate”,
sample mean is called the “estimator”.

Properties of a good Estimator:

1. An estimator should be unbiased

2. An estimator should be consistent
3. An estimator should be efficient
4. An estimator should be sufficient

Methods used for Point Estimation:

1. Method of maximum likelihood

2. Method of moments
3. Method of minimum variance
4. Method of least squares
5. Method of minimum chi-square
6. Method of inverse probability

Interval Estimation
An estimate which suggests the lowest and highest values within which population parameter
is expected to lie, they are called the interval estimates. Here, the two limits (lower and
upper) give an interval.

Qn: from the following data, find the limits within which population mean may lie:

Sample size = 100; Sample mean = 45; Sample S.D = 15

Sol:

x̄ = 45; S = 15; n = 100

 119

Here, since the sample is large, the test statistic = Z test

Z = Difference / SE
Degrees of freedom = infinity
The table value of Z at 5% level of significance and infinity d f = 1.96
Z = [x̄ ± μ] /[s/√n]
1.96 = [45 ± μ]/ [15/√100]
∴95% Confidence limits of population mean = 45 ± (1.96 x 1.5)
95% Confidence limits of population mean = 45 ± 2.94
95% Confidence limits of population mean = 42.06 and 47.94
Qn: Estimate the limits within which population mean lie at 95% level of confidence:
n = 25, x̄ = 4800, s = 500
Sol:
t = Difference / SE
Degrees of freedom = n-1 = 25-1 = 24
The table value of t at 5% level of significance and 24 d f = 2.064
t = [x̄ ± μ] /[s/√n-1]
2.064 = [4800 ± μ]/ [500/√25-1]
2.064 = [4800 ± μ]/ [500/√24]
2.064 = [4800 ± μ]/ 102.06
∴95% Confidence limits of population mean = 4800 ± (2.064 x 102.06)
95% Confidence limits of population mean = 4800 ± 210.65
95% Confidence limits of population mean = 4589.35 and 5010.65
Qn: Out of a sample of 500 items drawn from a population, 2% were found to be defective.
Estimate the proportion of population defectives at 95% confidence level. Also find number
of expected defectives in the daily production of 60,000 units.
Sol:
Sample proportion of defectives (p) = 2% = 0.02
Confidence level = 95%
Degrees of Freedom = infinity
95% Confidence limits of population proportion = p ± (Z x SE)
Here, SE = √(pq/n) = √(0.02 x 0.98)/500
= √0.0196/500 = 0.006261
95% Confidence limits of population proportion = 0.02 ± (1.96 x 0.006261)
= 0.02 ± 0.012272
 120

= 0.007728 and 0.032272

Expected number of defectives = 0.007728x60000; 0.032272 x 60000
= (463.68 , 1936.32)
= (464, 1936)
REVIEW QUESTIONS:

1. What do you mean by statistical estimation?

2. What are the two types of estimation?

3. What do you mean by Point Estimation?

4. What are the various methods used for point estimation?

5. What is meant by Interval Estimation?

6. Distinguish between Estimate and Estimator.

7. What are the important characteristics (properties) of a good estimator?

8. Distinguish between point estimation and interval estimation.

9. A random sample of 50 people from a population showed incomes with a mean =

50000 and Standard Deviation = 6000. Estimate the population mean with 95% and
99% confidence level.

10. In a sample of 500 units of a commodity from a large consignment, 40 units were
considered defective. Estimate the percentage of defective in the whole consignment
and limits within which the percentage will probably lie.

.*********.
 121

Chapter 16

SOFTWARES FOR QUANTITATIVE METHODS

Microsoft Excel: An Introduction
The aim of this chapter is to provide an introduction to using Microsoft Excel for
quantitative data analysis within the context of a business and management research project.
It covers some of the key features of Excel that are particularly useful when doing a research
project. Further it gives information on the use of Excel to apply various analysis techniques
discussed in various chapters. The information are presented here on the assumption that you
are already familiar with the basics of using Excel such as how to create worksheets, enter
data, use of formulae and functions, create charts (graphs), print and work, etc. If you have
never used Excel, there are many textbooks to get you started. Alternatively, you may find
that Excel training or support material is available in your institution. There are also various
websites, including Microsoft’s Office Support area (http://office.microsoft.com/en-
001/support/?CTT=97) that offers advice to get you started.

Why use Excel?

With so many specialist software packages available, why use Excel for statistical
analysis? Convenience and cost are two important reasons: many of us have access to Excel
on our own computers and do not need to source and invest in other software. Another
benefit, particularly for those new to data analysis, is to remove the need to learn a software
program as well as getting to grips with the analysis techniques. Excel also integrates easily
into other Microsoft Office software products which can be helpful when preparing reports or
presentations.

What you can do with Excel?

As a spreadsheet, Excel can be used for data entry, manipulation and presentation but
it also offers a suite of statistical analysis functions and other tools that can be used to run
descriptive statistics and to perform several different and useful inferential statistical tests that
are widely used in business and management research. In addition, it provides all of the
standard spreadsheet functionality, which makes it useful for other analysis and data
manipulation tasks, including generating graphical and other presentation formats. Finally,
even if using customised statistical software, Excel can be helpful when preparing data for
analysis in those packages.

Limitations of Excel
Even though it has wide applications and usage in data analysis, Excel is not free from
limitations. It remains first and foremost a spreadsheet package. Inevitably it does not cover
many of the more advanced statistical techniques that are used in research. More surprisingly,
it lacks some common tools (such as box plots) that are widely taught in basic statistics.
There is also concern amongst some statisticians over the format of specific output in some
 122

functions. The extensive range of graph (chart) templates is also criticised for encouraging
bad practice in data presentation through inappropriate use of colour, 3-D display, etc.
Despite these limitations Excel remains a very valuable tool for quantitative data analysis as
you will see.

Quantitative data analysis tools in Excel

Excel includes a large number of tools that can be used for general data analysis.
Here our primary concern is those that are relevant to the statistical and related analysis
techniques introduced in earlier chapters. Four sets of tools are particularly useful:

(1) Statistical functions:

Excel offers a broad range of built-in statistical functions. These are used to carry out
specific data manipulation tasks, including statistical tests. An example is the AVERAGE 1
function that calculates the arithmetic mean of the cells in a specified range. A list of Excel
functions referred to in this and other guides is included in Appendix A along with
instructions on how to access them.

(2) Data Analysis Tool Pak:

The Data Analysis Tool Pak is an Excel add-in. It contains more extensive functions,
including some useful inferential statistical tests. An example is the Descriptive Statistics
routine that will generate a whole range of useful statistics in one go. An introduction to
loading and using the Tool Pak add-in is included at Appendix B. The ToolPak is not
available in Excel for Mac. See Appendix B for an alternative.

(3) Charts:

Excel’s in-built charts (graphs) cover most of the chart types introduced in Chapter 13
and are invaluable in data exploration and presentation. We illustrate their use in Chapter 13
and also in the other guides.

(4) Pivot tables

Pivot tables provide a way of generating summaries of your data and organising data
in ways that are more useful for particular tasks. They are extremely useful for creating
contingency tables, cross-tabulations and tables of means or other summary statistics. A brief
introduction to creating pivot tables is given in the guide Data exploration in Excel:
univariate analysis.

Preparing Excel for analysis

Before starting, check that your Data Analysis ToolPak has been loaded. Do this by
selecting the Data tab; the Data Analysis command should appear in Analysis group on the
right-hand side of the ribbon.
 123

Setting up your data for analysis

Typically there are two options for getting your data into Excel:

1. Import the data in a suitable format from, for example, an online survey tool.
2. Enter the data manually.
If you are going to enter your data manually use a single worksheet to hold all the
data in your dataset and set up the worksheet with variables (questions) as the columns and
the cases (e.g. respondents) as the rows. An individual cell, therefore, contains a respondent’s
answer to a specific question.

Allocate column headers

In the first row, give each column a simple, informative header that will be easy to
understand when entering data or reviewing output. Avoid just using question numbers (e.g.
Q1, Q2, etc.) as these can be confusing if you have a large number of questions. Instead, use
a simple naming system. A variable measuring customer satisfaction, for example, could be
headed CSat: easy to remember and not likely to be confused during analysis. Ensure each
header is unique (this will facilitate subsequent analysis and avoid confusion when
interpreting output).

Allocate each case a unique ID

If they do not have one already, allocate each case in the dataset a unique numerical
identifier (ID). The easiest way to do this is simply to number them consecutively from 1
through to n (where n is the number of cases). For clarity, it is best to put the ID as the first
column in the worksheet. Giving each respondent a unique ID aids in sorting and tracking
individual responses when (for example) cleaning the data or checking outliers. A simple,
consecutive number ID system also makes it easy to reorder the data if needed. If you are
transferring data from paper copies of a questionnaire, it is useful to write the ID number
onto the paper copy to make it easier to check any errors.

Entering your data

Once the spreadsheet is set up, simply enter the data into the appropriate cell as required.
Numerical data can be entered as numbers, other data, such as Likert scale data, may need to
be coded .With nominal data you have two options:
 124

฀ Enter the values as words (e.g. male/female), appropriately abbreviated if required

(e.g. m/f). Ensure you are consistent in spelling and format as Excel will treat each
variation as a different value.

฀ Enter the re-coded numerical values (e.g. 0/1 for male/female), ensuring you keep a
record in a code book (Chapter 13). A worksheet in the workbook is a useful place to
record details of your variables and to store your code book as shown in Figure 3.

Which to do depends on your analysis needs. Some tools in Excel (e.g. pivot tables) work
well with text and generate meaningful output but some analysis tasks may require
numerically coded data. If you are exporting your data to another software package, check
the format required by that package. In some cases, it may be helpful to have both formats.
You can do this by creating a copy of the column containing the original data, then selecting
the new column and using Home > Find & Select > Replace to replace the original values
with the new ones. Ensure you give the new column a unique header.

Importing data
If you are importing the data from another electronic file, check that the layout is suitable (i.e.
respondents as rows, variables as columns), add or modify variable names if required, add
respondent ID if needed and check that the data has imported correctly.

Managing your data

Once you have created your dataset, ensure that you back it up in a secure place, not on
your PC or laptop. If you make any changes to your master dataset, record those changes
and create a duplicate back-up.

Give files a meaningful name. It is also helpful to date them as this makes it easier to track
back if you need to do so. Worksheet tabs can also be named to help you manage your
data.

Preparing your data

Once your data are entered you can follow the steps in Chapter 13 to prepare your data for
analysis. If you need to carry out data transformation, such as recoding variables or
calculating summated scores, do so now. (Hint: you can use functions such as SUM and
 125

AVERAGE to help you with creating summated scales.) If you are creating new variables
during data transformation ensure they are given unique column headers.

Example statistical functions

Function name Description
AVERAGE Returns the arithmetic mean (average) of the given numbers
CHISQ.DIST.RT Returns the right-tailed probability for the chi squared distribution
CHISQ.TEST Returns the p-value for the chi-squared test of association
CONFIDENCE.T Returns the margin of error for a confidence interval for the mean
COUNT Counts the number of cells in a range that contain numbers
COUNTIF Counts the number of cells in a range that meet a given condition
KURT Returns the kurtosis of a dataset
MAX Returns the maximum value of the given numbers
MEDIAN Returns the median of the given numbers
MIN Returns the minimum value of the given numbers
MODE.SNGL Returns the mode of the given numbers
PEARSON Returns the Pearson correlation coefficient (r) of two variables
SKEW Returns the skewness of a dataset
Returns the standard deviation of the given numbers, based on the
STDEV.P population
STDEV.S Returns the standard deviation of the given numbers, based on a sample
VAR.P Returns the variance of the given numbers, based on the population
VAR.S Returns the variance of the given numbers, based on a sample

Using a function
We will introduce specific functions in the other guides but the following example of
applying the AVERAGE function to calculate the mean age in the sample dataset in Figure
2 illustrates their use:

฀ Select the cell in which you wish the calculation to be placed (Hint: if you are
using the same worksheet as your dataset, avoid cells that are immediately
adjacent to your data).
 126

฀ Select Formulas > More Functions > Statistical > AVERAGE to open the
Function Argument dialogue box

SPSS (Statistical Package for Social Sciences)

SPSS (Statistical package for social sciences) is the set of software programs that are
combined together in a single package. The basic application of this program is to analyze
scientific data related with the social science. This data can be used for market research,
surveys, data mining, etc.

With the help of the obtained statistical information, researchers can easily understand
the demand for a product in the market, and can change their strategy accordingly. Basically,
SPSS first store and organize the provided data, then it compiles the data set to produce

suitable output. SPSS is designed in such a way that it can handle a large set of variable data
formats.

How SPSS Helps in Research & Data Analysis?

SPSS is revolutionary software mainly used by researchers which help them process
critical data in simple steps. Working on data is a complex and time consuming process, but
this software can easily handle and operate information with the help of some techniques.
These techniques are used to analyze, transform, and produce a characteristic pattern between
different data variables. In addition to it, the output can be obtained through graphical
representation so that a user can easily understand the result. Read below to understand the
factors that are responsible in the process of data handling and its execution.

1. Data Transformation: This technique is used to convert the format of the data. After
changing the data type, it integrates same type of data in one place and it becomes easy to
manage it. You can insert the different kind of data into SPSS and it will change its structure
 127

as per the system specification and requirement. It means that even if you change the operating
system, SPSS can still work on old data.

2. Regression Analysis: It is used to understand the relation between dependent and

interdependent variables that are stored in a data file. It also explains how a change in the
value of an interdependent variable can affect the dependent data. The primary need of
regression analysis is to understand the type of relationship between different variables.

3. ANOVA( Analysis of variance): It is a statistical approach to compare events, groups or

processes, and find out the difference between them. It can help you understand which method
is more suitable for executing a task. By looking at the result, you can find the feasibility and
effectiveness of the particular method.

4. MANOVA( Multivariate analysis of variance): This method is used to compare data of

random variables whose value is unknown. MANOVA technique can also be used to analyze
different types of population and what factors can affect their choices.

5. T-tests: It is used to understand the difference between two sample types, and researchers
apply this method to find out the difference in the interest of two kinds of groups. This test can
also understand if the produced output is meaningless or useful.

This software was developed in 1960, but later in 2009, IBM acquired it. They have made
some significant changes in the programming of SPSS and now it can perform many types of
research task in various fields. Due to this, the use of this software is extended to many
industries and organizations, such as marketing, health care, education, surveys, etc.

Advantages of SPSS:

The advantages of using SPSS as a software package compared to other are:

 128

• SPSS is comprehensive statistical software.

• Many complex statistical tests are available as a built in feature.

• Interpretation of results is relatively easy.

• Easily and quickly displays data tables.

• Can be expanded.

Limitations of SPSS:

Following are the important limitations of SPSS:

• SPSS can be expensive to purchase for students.

• Usually involves added training to completely exploit all the available features.

• The graph features are not as simple as of Microsoft Excel.

SPSS Windows and Files

SPSS statistics has three main windows and a menu bar at the top. These allow to:

(1) See your data

(2) See your statistical output
(3) See any programming commands you have written.

Each window corresponds to a separate type of SPSS file.

Students are directed to acquaint with the application of SPSS in performing testing of
hypotheses.

REVIEW QUESTIONS:

1. What is Microsoft Excel?

2. What are the important quantitative data analysis tools in Microsoft excel?
3. List the various statistical functions which can be performed in Microsoft excel.
4. What are the important limitations of Microsoft excel?
5. What is SPSS?
6. How does SPSS help in research and data analysis?
7. What are the advantages of SPSS?
8. What are the important limitations of SPSS?

.*********.